full-create-pr
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the repository to inform PR creation and code fixes, creating a surface for adversarial content to influence the agent.\n
- Ingestion points: Repository instructions, contribution documentation, PR templates, and Git hook outputs are read during the 'Inspect' and 'Commit and push' phases in
SKILL.md.\n - Boundary markers: The instructions lack explicit markers to differentiate between trusted system instructions and untrusted data from the repository files.\n
- Capability inventory: The skill is authorized to perform
gitoperations, modify local files to address linting issues, and use thelink_pull_requesttool.\n - Sanitization: There is no evidence of sanitization or filtering of the content retrieved from the repository before it is processed or included in outbound PR bodies.\n- [COMMAND_EXECUTION]: The skill uses shell commands to interact with the Git version control system, specifically for staging, committing, and pushing changes.\n- [DYNAMIC_EXECUTION]: The workflow requires the execution of Git hooks, which are local scripts that may be attacker-controlled if the repository itself is compromised. The skill specifically mandates that these hooks must not be bypassed.
Audit Metadata