The Agent Skills Directory

Indirect Prompt Injection (LOW): The skill best-practices-referencer processes data from multiple files in the references/ directory to create templates.
Ingestion points: Reads before-after-examples.md, prompt-patterns.md, common-workflows.md, anti-patterns.md, and best-practices-guide.md.
Boundary markers: Absent; no instructions provided to treat the reference content as non-executable data.
Capability inventory: Generates a 'Transformation Template' intended for use by a downstream transformation engine.
Sanitization: Absent.
Indirect Prompt Injection (LOW): The skill task-intent-analyzer processes user-supplied prompts to identify gaps and intent.
Ingestion points: Processes the user-provided 'original prompt'.
Boundary markers: Absent.
Capability inventory: Produces 'Transformation Guidance' and 'Interview Questions' that influence subsequent agent behavior.
Sanitization: Absent.

best-practices