The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from Git commits and pull requests to derive release scope and generate notes, creating a vector for instructions embedded in commit history to influence agent behavior.
Ingestion points: Commit messages and merged pull request content (SKILL.md).
Boundary markers: The skill does not define explicit delimiters or instructions to ignore potential commands within the ingested data.
Capability inventory: The agent can execute shell scripts, documentation tools, and the GitHub CLI, and can modify repository content (SKILL.md).
Sanitization: No sanitization or validation of the commit-based data is specified before it is processed by the agent.
[COMMAND_EXECUTION]: The skill is designed to run specific shell scripts and command-line utilities to perform verification and publishing steps.
Evidence: The workflow executes bash scripts/test-claude-plugin-install.sh and bash scripts/test-codex-plugin-install.sh (SKILL.md).
Evidence: The skill uses bun run docs:version and gh release create to manage documentation snapshots and GitHub releases (SKILL.md).

release