resolving-merge-conflicts
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to read external data including Git history, pull request descriptions, and issue trackers to gather context for conflict resolution. This constitutes an ingestion point for untrusted data that could potentially contain malicious instructions intended to influence the agent's resolution logic.
- Ingestion points: SKILL.md (Step 1 & 2: Git status, history, PRs, and issues).
- Boundary markers: None explicitly mentioned.
- Capability inventory: SKILL.md (Step 4 & 5: Shell execution for Git operations and project tests).
- Sanitization: None specified; the agent relies on its internal safety filters.
- [COMMAND_EXECUTION]: The skill directs the agent to identify and execute local automation commands such as type checkers, tests, and formatters defined within the project. This involves running code residing in the user's workspace to verify the merge.
- Evidence: SKILL.md (Step 4: 'Find the project's check commands and run them').
Audit Metadata