The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies and processes information from various project files (such as package.json, Makefile, and files within .claude/agents/ and .claude/skills/) to inform the planning process. This creates a surface for indirect instructions to influence the generated plan.
Ingestion points: The skill reads the Context Brief file and searches for existing verification scripts and project-specific agents/skills in the local directory.
Boundary markers: The instructions do not define explicit boundary markers or delimiters for data ingested from the project environment.
Capability inventory: The skill is restricted to file system read access for context discovery and file system write access for saving the implementation plan in the docs/ directory. It does not perform network operations or subprocess execution.
Sanitization: No specific sanitization or filtering logic is applied to the data gathered from the local project files before it is used to populate the plan document.

plan-crafting