The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs extensive shell command execution to manage the release lifecycle. This includes using git for version control, various build tools (npm, cargo, make, python -m build), and publishing utilities like twine and gh CLI. The execution happens autonomously in the default mode.
[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it ingests untrusted data from external sources.
Ingestion points: Reads git commit history via git log and GitHub metadata via gh pr list and gh issue list in agents/changelog-generator.md and SKILL.md.
Boundary markers: The skill lacks explicit 'ignore embedded instructions' delimiters when processing these external strings, relying instead on categorization logic within subagents to organize data before writing it to files.
Capability inventory: Possesses high-privilege capabilities including local file system modification, git push, and package publishing (npm publish, twine upload).
Sanitization: No specific sanitization or escaping of commit messages or PR content is mentioned before it is processed by the LLM agents.
[DATA_EXPOSURE]: The skill checks for the existence of sensitive credential files such as ~/.pypirc and environment variables like TWINE_PASSWORD to facilitate package publishing. While these are necessary for the skill's primary function, they involve the handling of sensitive authentication tokens.

ship