The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a wide range of local shell commands to manage development tasks. This includes executing pnpm for package management (pnpm install), running Vitest suites (pnpm test:changed), and performing code quality checks (pnpm check). It also utilizes specialized repo scripts like scripts/committer for git operations.
[COMMAND_EXECUTION]: It heavily relies on the GitHub CLI (gh) to interact with remote workflows. Commands such as gh run view --log and gh workflow run are used to manage CI status and trigger release validation processes.
[EXTERNAL_DOWNLOADS]: The skill is designed to fetch external data from GitHub, including CI job logs and E2E test artifacts. The pnpm test:docker:rerun utility specifically downloads summary.json and failures.json files from previous GitHub Action runs to generate targeted rerun commands.
[PROMPT_INJECTION]: The skill contains logic to process data from external GitHub logs and JSON artifacts, which represents an indirect prompt injection surface.
Ingestion points: GitHub Action logs fetched via gh run view --log and Docker E2E artifacts (summary.json).
Boundary markers: None present in the instructions for log analysis.
Capability inventory: The skill has access to shell execution (pnpm, gh, git) and file system writes for release evidence.
Sanitization: No explicit sanitization or validation of the ingested log content is described.
[SAFE]: The operations described are consistent with the intended purpose of a developer-oriented tool for CI/CD management. The skill explicitly instructs the agent not to store secrets in git and includes instructions for using configured environment tokens securely.

openclaw-testing