The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local build commands to verify project integrity.
Evidence: In scripts/verify_intake.py, the skill calls subprocess.run(["npm", "run", "build"], ...) to ensure the project remains buildable after documentation updates.
Context: This is a standard development workflow for a project-scaffolding and documentation tool.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-provided prompts and images, which represents a potential attack surface.
Ingestion points: scripts/load_intake.py reads external content from files provided via the --prompt-file argument. scripts/ingest_mockups.py ingest files from a user-specified --source-dir.
Boundary markers: Absent. There are no explicit delimiters or instructions to the model to ignore embedded commands within the ingested text.
Capability inventory: The skill has the ability to write files to the .heyeddi/ and designs/ directories and execute shell commands via npm.
Sanitization: The skill implements strict JSON schema validation in scripts/_product_schema.py and restricts mockup ingestion to a whitelist of image extensions (.png, .jpg, .jpeg, .webp, .svg) in scripts/ingest_mockups.py.

product-translator