agentsop-repo-map
Pass
Audited by Gen Agent Trust Hub on Sep 4, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill involves scanning a repository to build a context map for the agent. 1. Ingestion points: The agent processes source code from the repository's root directory (SKILL.md, SOP Phase 1). 2. Boundary markers: The skill explicitly distinguishes between the read-only 'repo-map' and the 'write set' to which the agent is authorized to make changes (SKILL.md, Section 2.3). 3. Capability inventory: The skill utilizes symbol extraction via tree-sitter and writes audit logs to the 'intermediate/' directory (SKILL.md, Operation 4). 4. Sanitization: The skill prioritizes symbol signatures and skeletons over raw code snippets, which naturally limits the amount of untrusted content directly influencing the prompt (SKILL.md, Section 2.1).
Audit Metadata