ai-scaffold-release

Pass

Audited by Gen Agent Trust Hub on Nov 9, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository's git history to generate release notes and derive versions, creating a surface for indirect prompt injection.\n
  • Ingestion points: Commit subjects are read using git log --format=%s (Step 2 in SKILL.md, assets/release_cli.py). The skill also processes git tags (git tag --list) and remote URLs.\n
  • Boundary markers: There are no explicit delimiters or instructions to the agent to ignore embedded commands within the commit messages during prompt interpolation.\n
  • Capability inventory: The skill possesses extensive capabilities across its scripts, including file writing (Step 5, 7c), network operations via REST APIs (assets/github_release_cli.py, assets/gitlab_release_cli.py), and shell command execution via uv run and subprocess.run.\n
  • Sanitization: Risk is partially mitigated by using json.dumps for API payloads to prevent injection into structured data and using git-cliff for commit parsing.\n- [COMMAND_EXECUTION]: The provided Python scripts frequently use subprocess.run to execute shell commands, primarily git and git-cliff.\n
  • Evidence: assets/release_cli.py, assets/promote_cli.py, assets/github_release_cli.py, and scripts/history_probe_cli.py all invoke git via subprocess to manage repository state and history.\n
  • Dynamic Extension: assets/promote_cli.py includes logic to execute an optional user-provided gate script, scripts/promote-gates.sh, if it is found and marked executable in the target repository.\n- [EXTERNAL_DOWNLOADS]: The skill relies on external tools and libraries that are fetched during installation or CI runs.\n
  • Evidence: assets/release_cli.py declares an inline dependency on git-cliff==2.10.1, which is managed by the uv tool. The CI templates (assets/gitlab-ci-release.yml, assets/github-release.yml) include instructions to install uv version 0.12.19 and specific Python interpreters.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 9, 2026, 02:43 PM