shunk031-manage-agent-guidance
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to process and organize instructions from various external sources, including repository-level configuration files (AGENTS.md, CLAUDE.md) and task-specific guidance.
- Ingestion points: The agent reads existing instructions from AGENTS.md, CLAUDE.md, and evaluates repository environment configuration through Makefile and pre-commit files.
- Boundary markers: The instructions do not specify the use of clear delimiters or warnings when the agent reads or interpolates content from these potentially untrusted sources.
- Capability inventory: The skill allows the agent to modify guidance files and execute repository gates (e.g., make, pre-commit, CI scripts).
- Sanitization: There is no mention of sanitizing or escaping the content retrieved from repository files before it is processed or used to influence agent behavior.
- [COMMAND_EXECUTION]: The skill directs the agent to run repository-specific validation gates such as Makefile targets, pre-commit hooks, and CI configuration checks to verify instruction changes.
- Evidence: The "Repository Validation" section explicitly instructs to "Inspect the affected repository's current Makefile, pre-commit configuration, and CI before naming a gate" and "Run its documented static gates."
Audit Metadata