The Agent Skills Directory

Prompt Injection (LOW): Indirect Prompt Injection Surface. The skill implements a pipeline where an LLM synthesizes release notes from git commit messages and updates GitHub Releases.
Ingestion points: Commit messages extracted via git log and responses from the OpenRouter/Gemini API.
Boundary markers: Absent. The instructions do not mandate the use of delimiters or 'ignore embedded instructions' markers when passing commit history to the LLM.
Capability inventory: The generated synthesize-release-notes.mjs script possesses the capability to write to the GitHub repository's releases using GITHUB_TOKEN permissions.
Sanitization: Absent. There is no requirement to sanitize or filter commit messages before processing, potentially allowing a malicious commit to influence the release notes content.
Command Execution (LOW): The skill generates a custom script (synthesize-release-notes.mjs) and configures GitHub Actions to execute it. While the script's purpose is functional, the dynamic generation and execution of code in CI/CD environments is a surface that requires careful agent implementation.
External Downloads (SAFE): The skill utilizes standard, reputable Node.js packages (e.g., semantic-release, commitlint) from the public npm registry. No untrusted or obfuscated external sources were identified.
Credentials Unsafe (SAFE): The skill follows best practices by instructing the user to store API keys and tokens (like GEMINI_API_KEY and GITHUB_TOKEN) in GitHub Secrets rather than hardcoding them in scripts or workflows.

changelog