greptile
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from an external source (GitHub PR comments) and uses that data to guide code modifications.
- Ingestion points: Fetches inline comments and review summaries using
gh apiin Step 2 ofSKILL.md. - Boundary markers: There are no explicit delimiters or instructions to treat the comment text as data rather than instructions when applying fixes.
- Capability inventory: The agent has the ability to read files, write files (Step 5), and execute shell commands via
ghandgit(Steps 1, 2). - Sanitization: No sanitization or validation of the comment content is performed before the agent evaluates and applies the suggested fixes, relying solely on the model's internal judgement.
- [COMMAND_EXECUTION]: The skill executes several shell commands to interact with the local environment and GitHub service.
- Evidence: Execution of
git branch --show-current,gh pr list,gh repo view, andgh apito gather context and metadata about the project and PR.
Audit Metadata