skill-library-mission
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze potentially untrusted project data—including READMEs, git history, CI configurations, and TODO/FIXME comments—to generate documentation and instructions. This creates a surface for indirect prompt injection if the repository source contains malicious instructions intended to subvert the generated skills.
- Ingestion points: The 'Discover' phase (Phase 1) investigates the entire file system, git history, and issue-shaped artifacts.
- Boundary markers: Absent; the instructions do not specify the use of delimiters or specific protection markers when the agent incorporates discovered text into new skill prompts.
- Capability inventory: The skill allows writing new files to the
.claude/skills/directory and executing local tools such aslsandrg. - Sanitization: Absent; the protocol relies on the model's own 'ground truth' verification against the repository rather than programmatic sanitization of the ingested content.
- [DYNAMIC_EXECUTION]: The skill allows for the creation of new executable content by directing generating agents to author diagnostic scripts based on discovered project patterns.
- Evidence: The taxonomy for the 'diagnostics-and-tooling' skill (Category 9) instructs the agent to 'ship actual scripts inside the skill's scripts/ dir where they exist or where you can write them.'
Audit Metadata