file-pr
Pass
Audited by Gen Agent Trust Hub on Aug 14, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, untrusted data from the local repository and existing GitHub metadata.\n
- Ingestion points: Content is retrieved from the outputs of
git log,git diff, andgh pr view(specifically the PR title and body).\n - Boundary markers: The instructions do not provide delimiters or boundary markers to isolate the ingested data, nor do they instruct the agent to disregard instructions potentially embedded in the git history.\n
- Capability inventory: The skill has the capability to write local files and execute shell commands through the agent's tools for git (add, commit, push) and the GitHub CLI (pr create, pr edit).\n
- Sanitization: The skill lacks logic to sanitize, filter, or escape the content obtained from commit messages or previous pull request descriptions before using it to generate a new pull request body.\n- [COMMAND_EXECUTION]: The skill explicitly uses multiple shell commands via
gitandghto perform repository operations. While standard for developer tooling, this represents a direct execution of commands based on the agent's processing of the workspace.
Audit Metadata