draft-pr

Pass

Audited by Gen Agent Trust Hub on Jul 11, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on local shell commands (git, gh, mktemp) to perform its functions. It uses string interpolation to build command arguments from dynamic content like branch names, PR titles, and bodies, which can be a surface for shell injection if the underlying platform does not sanitize these inputs.
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it ingests data from external sources, including existing PR content and repository configuration.
  • Ingestion points: PR titles, bodies, and base branch references are retrieved via gh pr view and gh repo view. Repository PR templates and Git diffs are also read into the agent context.
  • Boundary markers: The instructions do not define explicit delimiters or 'ignore' instructions to separate retrieved repository content from the agent's core logic.
  • Capability inventory: The skill possesses the ability to execute shell commands (git, gh), write to the local filesystem via redirection (cat > "$tmp_pr_body"), and push modifications to remote repositories.
  • Sanitization: There is no evidence of sanitization, escaping, or validation of the ingested data before it is interpolated into the agent's prompts or shell commands.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 11, 2026, 11:32 AM
Security Audit — agent-trust-hub — draft-pr