release-publisher

Pass

Audited by Gen Agent Trust Hub on Oct 28, 2026

Risk Level: SAFECOMMAND_EXECUTIONDATA_EXFILTRATIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from git commit history to summarize version changes and generate release notes. This creates an attack surface where malicious commit messages could influence the agent's summarization or inject instructions into public changelogs.
  • Ingestion points: Git commit history retrieved via git log or similar commands (referenced in Section 4).
  • Boundary markers: Absent. There are no instructions to the agent to treat commit content as data or to ignore instructions embedded within commits.
  • Capability inventory: The skill has permissions to write to local files (changelog.md, version.json, pyproject.toml), push to remote repositories (git push), and create/edit GitHub Releases (gh release create/edit).
  • Sanitization: No sanitization or escaping of commit content is mentioned before it is incorporated into the release notes.
  • [COMMAND_EXECUTION]: The skill performs extensive shell operations to manage the release process. This includes:
  • GitHub CLI: gh auth status, gh release view, gh release create, gh release edit, gh run list, and gh run view.
  • Git commands: git status, git tag, git ls-remote, git grep, git diff, git commit, and git push.
  • Local Scripts: Execution of PowerShell test scripts (.\tests\scripts\run-tests.ps1) and Python validation scripts (package/desktop/scripts/validate_components.py).
  • Build Tools: Execution of pnpm docs:build in the website directory.
  • [DATA_EXFILTRATION]: The skill includes a step to download release assets to a hardcoded external network path: \\DXP4800PLUS-72A\personal_folder\电脑软件\TrailSnap\X.Y.Z. While intended for backup, this represents a network operation to a non-whitelisted host and hardcodes internal network infrastructure details.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 28, 2026, 02:51 AM