skills/qzaixh/zhiheng/dl-deliver/Gen Agent Trust Hub

dl-deliver

Pass

Audited by Gen Agent Trust Hub on Oct 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill instructs the agent to perform repository management tasks using standard command-line tools.\n
  • Executes git commit and git push to deliver code changes to remote repositories.\n
  • Utilizes gh pr create and gh pr view to manage GitHub Pull Requests for code review.\n
  • Performs workspace cleanup using git worktree remove and file unlinking (unlink).\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external state and evidence files which serves as an ingestion point for potentially untrusted data.\n
  • Ingestion points: Reads task status and evidence from ../double-loop/references/runtime.md and ../double-loop/references/dependencies.md to determine task completion and PR descriptions.\n
  • Boundary markers: The instructions do not define specific delimiters or ignore instructions for the data being read from these external files.\n
  • Capability inventory: The skill possesses the ability to modify the remote repository via push, interact with the GitHub API for PR creation, and delete local files.\n
  • Sanitization: Employs SHA-256 content verification to ensure files have not changed before deletion, which provides a layer of integrity checking for file system operations.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 7, 2026, 05:46 AM