The Agent Skills Directory

[COMMAND_EXECUTION]: The skill invokes local utility scripts including python3 harness/scripts/unknown-ledger.py for tracking assumptions and python3 harness/scripts/code-logger.py for task ID generation. It also executes a Node.js tool located at ~/.agents/skills/archify/bin/archify.mjs to render sequence diagrams. These commands are used for internal auditing, logging, and documentation rendering within the agent's workspace environment.
[DYNAMIC_EXECUTION]: The skill dynamically generates sequence diagram specifications in JSON format based on the proposed plan and renders them using a shell loop. While this involves runtime command assembly, the process is constrained to generating documentation artifacts within the llmwiki/html/ directory.
[INDIRECT_PROMPT_INJECTION]: As a planning tool, the skill ingests user feature requests to generate specifications. It mitigates injection risks by:
Ingestion points: The initial user request is processed in the planning steps.
Boundary markers: Using highly structured Markdown sections (Context, Constraints, Plan, Requirements) to isolate data.
Capability inventory: File writes to specific wiki/html paths and execution of local logging/rendering scripts.
Sanitization: A mandatory human-in-the-loop approval step (Step 9) and a self-review phase (Step 8) ensure the generated plan is reviewed by the user before execution.

propose