The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes shell commands for environment maintenance and local data updates. Specifically, it uses pkill (Unix) and Stop-Process (Windows) to clean up orphaned development processes between waves. It also employs short node -e and python -c snippets to programmatically update feature_list.json with verification results. These actions are restricted to the local filesystem and the specific project context.
[PROMPT_INJECTION]: As an orchestrator, the skill reads untrusted repository content (e.g., source code, README files) to synthesize instructions for subagents. This creates a surface for indirect prompt injection. However, the skill incorporates mitigation strategies such as explicit boundary markers in prompt templates, mandatory verification steps, and a structured feature list to maintain control over the build process.

orchestrate