fix-merge-conflicts
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from source code files containing git conflict markers, creating a surface for indirect prompt injection. Attackers could place malicious instructions within these markers to influence the agent during the resolution process.
- Ingestion points: Conflicting files identified by git status in SKILL.md.
- Boundary markers: Absent. No delimiters are used to isolate conflicting code from instructions.
- Capability inventory: Execution of package managers, compilation, linting, and testing as described in SKILL.md.
- Sanitization: Absent. The agent is directed to process and validate the code directly.
- [COMMAND_EXECUTION]: The workflow involves executing local system commands for dependency management and build verification. While these are necessary for the skill's functionality, they involve running code derived from potentially untrusted merged files.
Audit Metadata