The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Node.js scripts (scripts/release-notes.mjs) to inventory Git history, record contributions, validate checkpoints, and render release notes. It also utilizes gh pr edit to update pull request descriptions.
[EXTERNAL_DOWNLOADS]: Fetches repository history, tags, pull request metadata, and issue data from the morluto/rea repository on GitHub.
[INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted content from the repository's history that could contain hidden instructions designed to influence the agent's behavior during the changelog rewriting process.
Ingestion points: Pull request bodies, commit messages (including co-author trailers), and issue reporter metadata ingested via the inventory command in SKILL.md.
Boundary markers: No explicit delimiters or specific instructions are provided to the agent to treat the ingested commit/PR data as untrusted text during the prose rewriting phase.
Capability inventory: The agent has the ability to execute shell commands (node, gh) and write files to the /tmp directory and the docs/releases/ path.
Sanitization: The skill mentions an "existing checkpoint validator" and a "new checker" used during the check and render phases, which validate the manifest structure and character limits, but their capacity to filter malicious natural language instructions is not defined.

rea-changelog-update