The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill summarizes untrusted data from external sources which could be used to inject instructions.
Ingestion points: Pull requests, git commit history, and issue tickets accessed via search_code and read_file as described in SKILL.md.
Boundary markers: The skill explicitly instructs that "automated changelog drafts are drafts only — require human editorial review" and that drafts should be treated as "untrusted content".
Capability inventory: The skill possesses the ability to create and modify files using write_file, edit_file, and create_file.
Sanitization: Mitigation strategies include a mandatory human editorial sign-off, secret scanning for credentials, and PII redaction before publication.

release-notes