gs-pull-request
Pass
Audited by Gen Agent Trust Hub on Nov 20, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses several command-line tools to interact with the local repository and GitHub. Specifically:
- Git Commands: It executes
git status,git branch,git symbolic-ref,git fetch,git log,git diff,git rev-parse,git merge-base, andgit pushto collect context and update the remote repository. - GitHub CLI (gh): It uses
gh pr view,gh pr create, andgh pr editto manage pull requests, including the use of--attachfor file uploads and--body-fileto safely pass content without shell interpretation risks. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository to generate its output, which is then pushed to a remote platform (GitHub).
- Ingestion points: Data is ingested from
git diffoutputs,git logmessages,GLOSSARY.mdfiles, and verification artifacts located in the.verify/directory. - Boundary markers: The skill provides instructions to ensure the agent only cites real file content and does not invent data, though it lacks explicit syntactic delimiters for the ingested content within the prompt context.
- Capability inventory: The skill possesses the ability to perform network operations (
git push) and create/edit remote resources (gh pr create/edit) based on the processed data. - Sanitization: The instructions explicitly mandate a review step where the agent is told to remove secrets, tokens, absolute local paths, and private details from both prose and screenshots before finalization.
Audit Metadata