resolving-merge-conflicts
Pass
Audited by Gen Agent Trust Hub on Sep 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to ingest and analyze external, untrusted content including commit messages, pull request descriptions, and issue trackers to determine the intent of code changes.
- Ingestion points:
SKILL.md(Step 2: Read commit messages, PRs, and issues). - Boundary markers: Absent; the instructions do not provide delimiters or warnings for the agent to ignore potentially malicious instructions embedded in the logs.
- Capability inventory: The skill requires the agent to perform git operations (commit, stage, rebase continue) and execute project-specific automated checks (tests, typecheckers) as described in
SKILL.md(Steps 4 and 5). - Sanitization: Absent.
- [COMMAND_EXECUTION]: The skill directs the agent to discover and run local project verification tools such as test suites and formatters, which involves the execution of shell commands within the project environment.
Audit Metadata