The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and executes data from the repository being processed, creating a surface for malicious content within that repository to influence the agent's actions.
Ingestion points: The skill reads and discovers commands from package.json, Makefile, Cargo.toml, pyproject.toml, and go.mod. It also uses git log --oneline -10 to infer changelog text for new releases.
Boundary markers: No explicit delimiters or instructions are provided to separate potentially untrusted content (like commit messages or script definitions) from the agent's core instructions.
Capability inventory: The skill has broad capabilities including shell command execution (npm run, npx), code modification (cat > file), and repository management (git push, gh pr merge).
Sanitization: The instructions do not mandate sanitizing or validating scripts found in project configuration files before they are executed in the local environment.
[COMMAND_EXECUTION]: The skill relies on several command-line tools to interact with version control systems and registries.
Evidence: Use of gh for pull request and CI management, git for staging and pushing release commits, and npm/npx for executing project gates and publishing.
[DYNAMIC_EXECUTION]: The skill uses runtime evaluation to generate dynamic identifiers.
Evidence: In references/changeset-and-commit.md, the skill uses node -e to execute a JavaScript snippet that generates a random alphanumeric string for use in changeset filenames.

autoship