The Agent Skills Directory

[SAFE]: The skill implements release management logic using localized Python scripts (semver.py, audit_changelog.py) and Markdown documentation. All logic is self-contained and operates on the local repository context.
[COMMAND_EXECUTION]: The script scripts/semver.py executes git tag --list via subprocess.run to retrieve version tags. This is a legitimate and necessary operation for the skill's primary purpose of calculating version bumps. The command is invoked without a shell and uses static arguments, posing no injection risk.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest data from external sources such as git log and public API diffs to generate changelog entries. While this data is technically untrusted, the risk is mitigated by the skill's design:
Ingestion points: Data is pulled from git log and git diff into the agent's context to facilitate changelog writing.
Boundary markers: The instructions rely on the agent's summarization capabilities to process these logs.
Capability inventory: The skill's capabilities are limited to local file modifications (CHANGELOG.md, pyproject.toml) and git metadata retrieval. No network access or arbitrary code execution paths are exposed to the ingested data.
Sanitization: The agent is explicitly instructed to transform raw commit subjects into user-facing descriptions, which serves as a manual sanitization step.

manage-releases