Bawwab: Top 302 scan

ponytail

Flagged as possibly malicious Reviewed: expected behaviour

Bawwab's static rules matched a pattern it treats as critical. Read the review notes: a reviewed false alarm is marked as such.

Why it was flagged

The static rules marked this “Flagged as possibly malicious” because of:

Kind
Plugin
Source
github
Repository
DietrichGebert/ponytail
Commit
96608e078b2d5ea607c5c3b8b5a54d1c1fba6846
Listed in
DietrichGebert/ponytail
Popularity
195,197 GitHub stars (2026-11-20)
Link
https://github.com/DietrichGebert/ponytail/tree/96608e078b2d5ea607c5c3b8b5a54d1c1fba6846
Scanned
2026-11-20T22:42:19+00:00
Scan time
7.8 s
Scanned as
plugin
Review note: A coding-style plugin whose only outbound call is an opt-in, user-configured webhook (off by default); the scanner hits are benchmark scripts, tests and PNG images.

Findings (12)

files.truncated: Scan limit reached med

assets/codex-icon.png

assets/codex-icon.png contains NUL bytes; not scanned

What this means: Some files, or the tail of a very large file, were not scanned, so malicious code could hide in them.

The rule's general advice: Review the unscanned files manually or raise the limits. (see the review note for this case)

Known false alarm

A PNG icon; binary image data (NUL bytes are normal for PNG), not code.

Reviewed by Bawwab maintainers, 2026-11-20

files.truncated: Scan limit reached med

assets/logo.png

assets/logo.png contains NUL bytes; not scanned

What this means: Some files, or the tail of a very large file, were not scanned, so malicious code could hide in them.

The rule's general advice: Review the unscanned files manually or raise the limits. (see the review note for this case)

Known false alarm

A PNG logo image; binary image data (NUL bytes are normal for PNG), not code.

Reviewed by Bawwab maintainers, 2026-11-20

credential-access: Touches credentials or secrets high

.claude-plugin/plugin.json:31

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": "Optional: sent to the webhook as a Bearer token. Kept in the system keychain, never in a settings file or a repo."

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

A config-field description for an optional webhook token. The webhook is empty by default; if the user sets a URL, hooks/band/metrics.ts sends the task's metrics row as JSON to it, and refuses plain http unless the target is local.

Reviewed by Bawwab maintainers, 2026-11-20

credential-access: Touches credentials or secrets high

benchmarks/agentic/complete.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

Judge: --model (default claude-sonnet-4-6). API key from ../../.env (shared with judge.py), or 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)

Expected for what this tool does

Docstring in a developer benchmark script saying the LLM judge reads an Anthropic API key from a local .env or the environment. It is not part of the plugin's runtime hooks.

Reviewed by Bawwab maintainers, 2026-11-20

credential-access: Touches credentials or secrets high

benchmarks/agentic/judge.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

Judge: claude-sonnet-4-6 via the Anthropic Messages API (key from ../../.env). Scores the SOURCE

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

Docstring in the benchmark judge. load_key() reads ANTHROPIC_API_KEY from a local .env or the environment and sends it only as the x-api-key header to api.anthropic.com, or falls back to the claude CLI.

Reviewed by Bawwab maintainers, 2026-11-20

exfil.combo: Reads credentials and makes network calls med

benchmarks/agentic/judge.py:13

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

Judge: claude-sonnet-4-6 via the Anthropic Messages API (key from ../../.env). Scores the SOURCE

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

The benchmark judge reads an Anthropic key and posts to api.anthropic.com, the key's own provider. The only other network call in the file is that same API; nothing is sent to a third party.

Reviewed by Bawwab maintainers, 2026-11-20

credential-access: Touches credentials or secrets high

benchmarks/agentic/tasks.py:543

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, "PORT": str(port)}

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

Copies the environment to start the benchmark's own generated node server on a random port; the harness then talks to 127.0.0.1 only.

Reviewed by Bawwab maintainers, 2026-11-20

exfil.combo: Reads credentials and makes network calls critical

benchmarks/agentic/tasks.py:543

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 = {**os.environ, "PORT": str(port)}

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

The environment copy is passed to a local child process, and the only HTTP requests go to 127.0.0.1. Nothing leaves the machine.

Reviewed by Bawwab maintainers, 2026-11-20

credential-access: Touches credentials or secrets high

hooks/band/register.tsx:83

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

const home = async ($: EngineInterface) => (await $.env.get('HOME')) ?? (await $.env.get('USERPROFILE'))

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

Reads HOME or USERPROFILE to locate the plugin's own '.ponytail-band' flag file under the Claude config directory. No credentials are read.

Reviewed by Bawwab maintainers, 2026-11-20

credential-access: Touches credentials or secrets high

tests/band.test.tsx:294

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

on('fs.read', ($, e, next) => (e.path === '/home/u/.claude/.ponytail-band' ? { value: 'off' } : next(e)))

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

Test code that mocks a read of the plugin's own flag file at a fake /home/u path.

Reviewed by Bawwab maintainers, 2026-11-20

credential-access: Touches credentials or secrets high

tests/hooks-windows.test.js:117

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

const env = { ...process.env, HOME: home, USERPROFILE: home, CLAUDE_CONFIG_DIR: path.join(home, '.claude'), PONYTAIL_MAP: '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

Test code that points HOME at a temporary directory before running the hook scripts.

Reviewed by Bawwab maintainers, 2026-11-20

credential-access: Touches credentials or secrets high

tests/hooks.test.js:65

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

statePath: path.join(home, '.claude', '.ponytail-active'),

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

Test code that checks the path of the plugin's own '.ponytail-active' state file in a temporary home.

Reviewed by Bawwab maintainers, 2026-11-20

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

files.binary-unscanned: Binary files not analysed low

1 binary file not analysed; use --deep to run in the sandbox

What this means: Prebuilt executables cannot be read by static rules, so what they do is unknown until they run.

What you can do: Use --deep to run the package in the sandbox, or prefer packages built from source.

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

benchmarks/agentic/run.py:54

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

PLUGIN_CACHE = Path.home() / ".claude" / "plugins" / "cache"

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

hooks/ponytail-config.js:73

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

return process.env.CLAUDE_CONFIG_DIR || path.join(os.homedir(), '.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

hooks/ponytail-statusline.ps1:2

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

$ClaudeDir = if ($env:CLAUDE_CONFIG_DIR) { $env:CLAUDE_CONFIG_DIR } else { Join-Path $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/band-import.test.js:93

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

const withCwd = r => ({ ...r, cwd: '/work/shop/.claude/worktrees/brave-otter' });

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/band.test.tsx:428

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

return reads.at(-1)?.endsWith('C:/Users/u/.claude/.ponytail-band') ? { value: 'off' } : next(e)

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/cursor-hooks.test.js:135

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

assert.equal(fs.existsSync(path.join(c.home, '.claude')), false, 'Cursor state must not land in ~/.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/hooks.test.js:17

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

assert.equal(isShellSafe('/home/u/.claude/plugins/ponytail/hooks/ponytail-statusline.sh'), 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/uninstall.test.js:24

MITRE ATLAS AML.T0081 Modify AI Agent Configuration

const claudeDir = path.join(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.

Dispute this result

Maintainer and think this is wrong? Open a dispute on GitHub with the target id plugin-dietrichgebert-ponytail-1cb7b1.