amd-skills
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:
remote-code.pipe-shell: Pipes a download into a shell (high, 1× · first at tracelens-analysis-orchestrator/evals/eval_scripts/run_pi_analysis_in_docker.sh:511)
- Kind
- Plugin
- Source
- github
- Repository
- amd/skills / skills
- Commit
d4417422312c82d4761d3255b69bfcd411147cf3- Listed in
- anthropics/claude-plugins-official
- Popularity
- 508 GitHub stars (2026-11-20) (official, always included)
- Link
- https://github.com/amd/skills/tree/d4417422312c82d4761d3255b69bfcd411147cf3/skills
- Scanned
- 2026-11-20T23:27:02+00:00
- Scan time
- 3.4 s
- Scanned as
- plugin
Review note: A maintainer eval script and a template comment; nothing runs at install.
Findings (2)
remote-code.pipe-shell: Pipes a download into a shell high
tracelens-analysis-orchestrator/evals/eval_scripts/run_pi_analysis_in_docker.sh:511
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
curl -fsSL "https://deb.nodesource.com/setup_${NODE_MIN_MAJOR}.x" | bash -
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)
Expected for what this tool does
An evaluation script, run by hand, that sets up an inference-server Docker container and installs Node.js inside it from NodeSource's setup script. The curl | bash is real but runs only when a developer runs the eval, not at install.
Reviewed by Claude (AI-assisted review), 2026-11-21
injection.html-comment: Hidden instruction in an HTML comment med
tracelens-analysis-orchestrator/agents/kernel-fusion-analyzer.md:182
OWASP LLM LLM01 Prompt InjectionOWASP Agentic ASI01 Agent Goal HijackMITRE ATLAS AML.T0051.001 LLM Prompt Injection: IndirectMITRE ATLAS AML.T0068 LLM Prompt ObfuscationCWE CWE-1427 Improper Neutralization of Input Used for LLM Prompting
<!-- [comparative] Two Data tables with the same columns — you MUST include BOTH, each labelled with that trace's E2E ms. Do NOT add a totals line: -->
What this means: HTML comments are invisible when a document is rendered but are still read by the AI, so they can carry instructions meant to stay unseen.
The rule's general advice: Read the comment; remove it or do not install. (see the review note for this case)
Known false alarm
A formatting note in an output template (include both trace tables, no totals line). It shapes the report and asks for nothing hidden.
Reviewed by Claude (AI-assisted review), 2026-11-21
3 low-severity notes (informational, do not affect the result)
install-dir-access: Reads or writes Claude's install folders low
flydsl-kernel-authoring/.federated.json:6
MITRE ATLAS AML.T0081 Modify AI Agent Configuration
"path": ".claude/skills/flydsl-kernel-authoring",
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
flydsl-kernel-trace-analysis/.federated.json:6
MITRE ATLAS AML.T0081 Modify AI Agent Configuration
"path": ".claude/skills/kernel-trace-analysis",
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
moat-cuda-to-rocm/.federated.json:6
MITRE ATLAS AML.T0081 Modify AI Agent Configuration
"path": ".claude/skills/cuda-to-rocm",
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-amd-skills-skills-b1f8fe.