The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to use the GitHub CLI (gh) to list runs, view logs, and watch workflow progress. It also involves running standard build and test commands such as npm run lint, npm test, pytest, and pnpm test. These operations are consistent with the skill's stated purpose of CI/CD triage and utilize established developer tools within the local environment.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data in the form of GitHub Action logs. There is a theoretical surface for indirect prompt injection if log files contain malicious instructions intended to mislead the subagents during triage.
Ingestion points: Workflow logs are fetched via gh run view --log or pasted manually into the conversation as relevant log excerpts for subagents.
Boundary markers: The skill suggests providing log excerpts within clear sections, though it does not mandate specific sanitization markers beyond redacting secrets.
Capability inventory: The triage subagents have the capability to read/edit code and execute local build/test commands to verify fixes.
Sanitization: The skill advises redacting secrets before pasting logs but does not explicitly include instructions to filter or escape potential injection patterns embedded within logs.

parallel-ci-triage