lynx-repair
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill represents an indirect prompt injection attack surface because it ingests untrusted data from GitHub pull request comments via the
gh pr view --commentscommand. - Ingestion points: Pull request comments and review threads fetched in Phase 1 (
SKILL.md). - Boundary markers: The instructions lack explicit delimiters or "ignore embedded instructions" warnings for the agent when processing these comments.
- Capability inventory: The agent has the capability to modify local files, execute
git add,git commit, and use thegh apito post responses to the pull request thread (SKILL.mdPhases 4 and 5). - Sanitization: No specific sanitization or validation of the comment content is performed before the agent uses it to decide on code fixes or commit messages.
- [COMMAND_EXECUTION]: The skill utilizes several local command-line tools to perform its tasks.
- Evidence: Execution of
gh(GitHub CLI) to view PRs and post reviews,gitto stage and commit changes, and a custom CI toolgwto verify code changes. - [EXTERNAL_DOWNLOADS]: The skill invokes
pnpm installduring the verification phase to synchronize dependencies. - Evidence: The command
pnpm installis executed in theVERIFYphase to ensure the environment is ready for testing changes. This uses the well-known NPM registry service.
Audit Metadata