The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes extensive use of shell commands to orchestrate its workflow.
Evidence: Uses jq for parsing loop data, test for artifact validation, and npm test for integration verification.
Evidence: Employs node with heredocs to dynamically generate and execute JavaScript for packaging delivery artifacts in Step 9.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by design as a meta-orchestrator processing external data.
Ingestion points: Reads task descriptions and research from .claude/.artifacts/loop1-planning-package.json to generate the agent-skill-assignments.json matrix.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are used when interpolating external data into agent prompts.
Capability inventory: The skill can execute shell commands (npm, npx, node), write to the filesystem, and spawn new agent tasks with dynamically generated instructions.
Sanitization: No sanitization or validation of the external JSON content is performed before it drives the assignment of skills and instructions to other agents.
[EXTERNAL_DOWNLOADS]: Executes logic from an external package registry via npx.
Evidence: Calls npx claude-flow@alpha to perform memory operations and coordination. While the NPM registry is a well-known service, the package itself is an external dependency executed at runtime.

parallel-swarm-implementation