The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes git and gh (GitHub CLI) for repository management and pull request operations. It also executes a repository-local script .hooks/hard-eng.py to coordinate the shipping process.
[DYNAMIC_EXECUTION]: The skill supports deployment verification by executing commands defined in the hard-eng.gates.json configuration file. This allows for the execution of project-specific verification scripts (e.g., python3 scripts/verify_deployment.py) based on local configuration.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests data from local plan files and pull request metadata to generate deployment reports and PR content.
Ingestion points: PLAN.md, hard-eng.gates.json, and output from gh pr commands.
Boundary markers: The references/checks.md file includes a 'Publication privacy' section with guidelines for audience establishment and payload review.
Capability inventory: Executes shell commands (git, gh) and local Python scripts, and writes content to the local filesystem and remote pull requests.
Sanitization: The instructions explicitly require inspecting and sanitizing all outgoing payloads (diffs, plans, logs) to remove private identifiers and operational data before publication.

he-ship