wiki-researcher
Pass
Audited by Gen Agent Trust Hub on Feb 23, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes
git remote get-url originandgit rev-parse --abbrev-ref HEADto determine the repository's origin URL and current branch name for citation formatting. - [PROMPT_INJECTION]: The skill analyzes untrusted data from a codebase, which constitutes a surface for indirect prompt injection.
- Ingestion points: Local codebase files (as defined in SKILL.md for iterative research).
- Boundary markers: Absent; there are no explicit instructions for the agent to ignore or isolate instructions found within code comments or documentation.
- Capability inventory: File system read access and execution of local git commands.
- Sanitization: None; the instructions explicitly require the agent to read real implementations and follow all code paths without filtering.
Audit Metadata