The Agent Skills Directory

[SAFE]: The skill implements a 'propose-don't-apply' workflow, which prevents the automated execution of commands without explicit user consent. Evidence is found in the 'CRITICAL guardrails' section of SKILL.md.
[COMMAND_EXECUTION]: Uses the GitHub CLI (gh) to retrieve logs and PR information. All command patterns are standard for CI/CD workflows and rely on the user's existing environment permissions.
[PROMPT_INJECTION]: The skill processes external CI log data which presents a theoretical surface for Indirect Prompt Injection.
Ingestion points: Job logs fetched via gh api as described in SKILL.md (Step 2: Fetch the failing log).
Boundary markers: No explicit boundary markers are used for the raw log data.
Capability inventory: The skill utilizes Bash, Grep, and the gh CLI for its operations.
Sanitization: Risk is mitigated by a strict policy against automatic execution; all proposed fixes require explicit user verification and approval.

ci-debug