release

Pass

Audited by Gen Agent Trust Hub on Mar 23, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill contains explicit instructions for the agent to operate in "AUTONOMOUS MODE," specifically stating "Do NOT ask questions" and "Do NOT pause for confirmation." This directive is designed to bypass standard interaction loops, which reduces the user's opportunity to review or intercept actions during the automated configuration of release pipelines and CI workflows.
  • [COMMAND_EXECUTION]: The skill performs extensive system operations including package installation (npm install, pip install, cargo install), project state detection (git log, git tag, git remote), and tool verification (npx semantic-release --dry-run). It also writes telemetry data to a local project directory at ~/.claude/projects/.
  • [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it incorporates untrusted data from the user's environment into its logic and generated artifacts.
  • Ingestion points: The skill reads data from repository manifest files (package.json, pyproject.toml, etc.) and recent commit messages via git log --oneline -20 to detect versioning patterns and repository metadata.
  • Boundary markers: No delimiters or instructions are used to treat the ingested repository data as untrusted or to ignore any potential instructions embedded within commit messages or package descriptions.
  • Capability inventory: The skill has the capability to generate and write GitHub Action workflows (.github/workflows/release.yml) that possess elevated permissions (contents: write, pull-requests: write) and interact with sensitive repository secrets like GITHUB_TOKEN and NPM_TOKEN.
  • Sanitization: There is no evidence of sanitization or validation for the metadata extracted from the project environment before it is used to populate configuration templates or CI scripts.
  • [EXTERNAL_DOWNLOADS]: The skill automates the installation of several third-party utilities from official registries, including semantic-release, python-semantic-release, changesets, and cargo-release. These are well-known industry tools used for version management.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 23, 2026, 10:58 AM
Security Audit — agent-trust-hub — release