resolving-merge-conflicts
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill directs the agent to ingest and interpret metadata from external and potentially untrusted sources like git commit history and pull request descriptions.
- Ingestion points: Step 2 of SKILL.md requires reading commit messages, PRs, and original issues/tickets.
- Boundary markers: There are no explicit instructions or delimiters to isolate untrusted content from the agent's core instructions.
- Capability inventory: The skill workflow includes executing git commands and running arbitrary project-defined 'automated checks' (tests, formatters, etc.) in steps 1, 3, 4, and 5 of SKILL.md.
- Sanitization: No validation or sanitization of the retrieved external data is specified before it is processed by the agent.
Audit Metadata