configure-claude-rules
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill includes a workflow (Step 5) that offers to search the web for "best practices" and then "author a new markdown file" based on those results to be placed in the repository's rules directory.
- Ingestion points: External web search results are ingested into the agent's context to generate new rule files (SKILL.md).
- Boundary markers: The instructions do not specify any delimiters or warnings to the agent to ignore potentially malicious instructions embedded in the search results.
- Capability inventory: The skill has the capability to write files to the repository using
scripts/write-rules.py(SKILL.md, write-rules.py). - Sanitization: No explicit sanitization or validation of the search content is described before it is used to generate executable behavior rules.
- [COMMAND_EXECUTION]: The skill uses the
uvtool to execute local Python scripts for repository analysis and file management. - Evidence:
uv run scripts/detect-topics.py,uv run scripts/check-rules.py, anduv run scripts/write-rules.pyare executed to automate the rule configuration process (SKILL.md).
Audit Metadata