pr-bump
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external data by reading Pull Request titles, numbers, and contents via the GitHub CLI. This data is used to construct git commit messages and package manager commands. While the skill provides clear instructions to manually verify PR content and audit dependency sources, the ingestion of attacker-controlled PR data represents a potential attack surface.\n
- Ingestion points: GitHub CLI output (
gh pr list,gh pr view).\n - Boundary markers: Instructions explicitly require inspecting PR diffs, commits, and file lists before proceeding, and enforcing a 'supply-chain floor' by verifying dependency resolution.\n
- Capability inventory: The skill uses
git commit,git push,bundle update, andbin/deploy(SKILL.md).\n - Sanitization: The instructions focus on manual verification and auditing rather than technical sanitization or escaping of the ingested strings.\n- [COMMAND_EXECUTION]: The skill performs various command-line operations, including package management updates (
bundle update) and a local deployment script (bin/deploy). These operations are necessary for the skill's primary purpose. The instructions mitigate risks by requiring the agent to audit dependencies for suspicious hooks, typosquatting, or non-standard sources before execution.\n- [CREDENTIALS_UNSAFE]: The documentation includes hardcoded environment variables forACTIVE_RECORD_ENCRYPTION. These are dummy hex sequences (e.g.,0123456789abcdef...) used to provide deterministic test keys for CI-equivalent checks, and do not represent actual production secrets.
Audit Metadata