The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from user-supplied evidence and external tool outputs. Such data could potentially contain malicious instructions intended to influence the agent's behavior.
Ingestion points: The skill ingest alerts, logs, user reports, and output from the gh CLI (PRs and workflow runs) as evidence in SKILL.md.
Boundary markers: Instructions include separating observed facts from hypotheses and redacting sensitive information, which provides a logical boundary for processed data.
Capability inventory: The skill is capable of writing incident documentation to the local filesystem at <artifacts-root>/specs/incidents/.
Sanitization: Includes a rule to redact tokens, keys, passwords, and PII from evidence before it is included in output notes.
[EXTERNAL_DOWNLOADS]: Fetches pull request and deployment metadata from GitHub via the official gh CLI tool. This represents standard integration with a well-known service to provide context for incident timelines.

incident-triage