The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill generates interactive prototypes that reference library resources from well-known technology services.
Evidence: Templates in references/prototype-templates.md include script tags for unpkg.com, cdnjs, and jsdelivr to load Vue, React, and Babel.
[COMMAND_EXECUTION]: The skill automates a complex workflow by orchestrating the execution of several internal sub-skills.
Evidence: Sequential calls to /pm-ia, /pm-state, /pm-flow, and /pm-wireframe to generate various diagram types.
[PROMPT_INJECTION]: The skill ingests untrusted project documentation to derive its visualization outputs, creating an exposure surface for indirect prompt injection.
Ingestion points: Reads product context from docs/pm-context/pm-context.md.
Boundary markers: No explicit delimiters or instructions are used to ignore embedded commands within the ingested file.
Capability inventory: Ability to read file system metadata, write markdown and HTML files, and chain internal commands.
Sanitization: No sanitization or filtering of the source context is documented.
[SAFE]: The skill behavior is consistent with its primary purpose of generating design artifacts. All data access is restricted to non-sensitive project configuration files, and network operations are limited to referencing well-known public CDNs.

pm-sketch