The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. It instructs the agent to read git commit messages and file diffs to generate release summaries. Malicious content within these commits could potentially influence the agent's behavior during the summarization process.
Ingestion points: git log and git diff outputs are used in Phase 4 of the workflow to generate changelog entries.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the workflow description when processing git data.
Capability inventory: The skill executes subprocesses (git, uv, ./scripts/fix), writes to local files (CHANGELOG.md), and performs network operations (git push).
Sanitization: There is no evidence of sanitization or filtering of the commit message content before it is processed by the AI.
[COMMAND_EXECUTION]: The skill executes project-local scripts. The commit-and-push script specifically runs ./scripts/fix from the repository's root directory. This behavior executes code defined within the repository being operated on, which represents a risk if the repository itself is malicious.
[EXTERNAL_DOWNLOADS]: The skill performs automated dependency management using uv sync. This command fetches and installs Python packages from the public Python Package Index (PyPI), which is a standard external source.
[COMMAND_EXECUTION]: The skill relies extensively on system commands such as git and uv to manage the release lifecycle, including staging files, committing changes, tagging releases, and pushing code to remote servers.

release