The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several shell-based tools including git, uv, make, and the GitHub CLI (gh). These are used to calculate diffs, sync dependencies, run local validation scripts (scripts/check_version_tag.py, scripts/check_kimi_dependency_versions.py), and manage pull requests.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process data from the local repository which could potentially contain malicious instructions if the repository itself was compromised.
- Ingestion points: Data is read from git diff outputs, pyproject.toml, CHANGELOG.md, and breaking-changes.md (SKILL.md Step 2, 5).
- Boundary markers: None are present; the skill lacks specific instructions for the agent to ignore or delimit potentially untrusted content within these files.
- Capability inventory: The skill has the ability to write to the file system, execute shell commands (git, uv, make, gh), and open network-facing pull requests.
- Sanitization: The risk is mitigated by mandatory human-in-the-loop confirmation steps. The agent is instructed to wait for explicit user approval before version bumps (Step 3) and before opening the final PR (Step 9).

release