The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local setup scripts (setup-product-init.sh and setup-product-init.ps1) to resolve repository paths, detect feature areas, and calculate the next available document ID.
[INDIRECT_PROMPT_INJECTION]: The analysis pipeline ingests data from untrusted external sources, specifically the user's codebase, documentation (READMEs, ROADMAPs), and pricing configuration.
Ingestion points: Codebase directory structure, README.md, ROADMAP.md, pricing.md, and service definitions in docker-compose.yml via the bash and PowerShell setup scripts.
Boundary markers: The instructions do not define explicit delimiters or instructions to separate raw data from agent instructions during the Discovery, Pattern, or Synthesis phases.
Capability inventory: The skill facilitates reading files across the repository and writing generated Markdown files and a state JSON file to the .adlc/ directory.
Sanitization: The setup scripts employ standard JSON conversion utilities (ConvertTo-Json in PowerShell and json.dumps in a Python subprocess) to safely format metadata and detected areas.

product-init