The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from external sources and interpolates it into subagent prompts.
Ingestion points: Processes spec.md (Phase 1), output.yaml from subagents (Phase 4), and project configuration files like package.json and Makefile (Phase 1/5).
Boundary markers: The skill instructions do not specify the use of delimiters or 'ignore embedded instructions' warnings when injecting external content into subagent prompts.
Capability inventory: The skill possesses significant capabilities including the execution of arbitrary shell commands (build/test/lint), Git repository modifications, and spawning/managing other AI agents via the Task tool.
Sanitization: There is no evidence of content sanitization or validation of the external strings before they are used to generate task plans or instructions.
[COMMAND_EXECUTION]: The skill automatically detects and executes verification commands (build, test, lint) found within the user's project codebase.
Evidence: Phase 1 (Step 6) and Phase 5 explicitly instruct the agent to identify commands from package.json, Makefile, or other project documentation and run them to establish baselines or verify implementations.

dispatch