The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided architecture and PRD documents to generate feasibility reports. While this creates a theoretical surface for indirect prompt injection, the skill includes specific constraints, such as ignoring Mermaid diagram syntax for technology detection, to mitigate potential data-driven instruction overrides. \n- Ingestion points: architecture_doc_path and prd_path (SKILL.md, step-01-init.md) \n- Boundary markers: Present (explicitly excludes Mermaid diagrams from parsing logic in references/coverage-patterns.md) \n- Capability inventory: Spawns parallel subagents for API analysis; executes a local Python script for file writing (step-01, step-03, step-06) \n- Sanitization: Not explicitly implemented for the ingested text content.\n- [COMMAND_EXECUTION]: The skill uses a local script src/shared/scripts/skf-atomic-write.py to perform atomic file operations. This is a standard internal utility within the project structure and does not involve external command injection surfaces.

skf-verify-stack