shunk031-manage-agent-guidance
Pass
Audited by Gen Agent Trust Hub on Sep 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill's primary function is to organize and reorganize agent guidance files (like
AGENTS.mdandCLAUDE.md). These files are external to the skill and can be modified by any repository contributor, making them a vector for indirect prompt injection. \n - Ingestion points: The workflow involves reading higher-level instructions and repository-local guidance files (
SKILL.md, step 3). \n - Capability inventory: The skill references local binaries (
prek,codex) and the use of command-line tools like the GitHub CLI (gh). \n - Boundary markers: There are no defined delimiters to separate the data being managed from the operational instructions of the skill. \n
- Sanitization: The instructions do not include steps for sanitizing or validating the content of the guidance files before processing. \n- [COMMAND_EXECUTION]: The skill requires the agent to execute specific local command-line tools to validate the guidance files. \n
- Evidence: The
Repository Validationsection inSKILL.mdmentions runningprekand using acodexexecutable for local validation.
Audit Metadata