debrief
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute multiple git commands to identify branch names, find merge bases, and retrieve commit logs and diffs.
- Evidence: Commands include
git symbolic-ref,git merge-base,git rev-parse,git log,git diff, andgit status. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze external data from the local repository, which may include source code, configuration files, and git history. This content could contain malicious instructions meant to influence the agent's behavior or bias its analysis.
- Ingestion points: Source code files, repository manifests (e.g., README, package.json, pyproject.toml), and git output are read into the agent's context in SKILL.md.
- Boundary markers: The skill does not specify any delimiters or instructions for the agent to ignore instructions embedded within the codebase being analyzed.
- Capability inventory: The skill allows file reading and git command execution, but specifically prohibits file writes, staging, or commits ('Never edit files, stage, or commit').
- Sanitization: No explicit sanitization or filtering of the ingested code content is described.
Audit Metadata