The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill uses scripts/fetch_asset.py and scripts/render_product_shot.py to download assets and generated images from remote URLs. Specifically, it uses urllib.request to fetch brand logos and photos, and the fal-client library to interact with image generation services.
[COMMAND_EXECUTION]: The workflow relies on the execution of multiple Python scripts provided within the skill (scaffold_brand.py, fetch_asset.py, register_asset.py, render_product_shot.py, verify_pack.py). These scripts perform file system operations, project scaffolding, and asset registration.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) because it ingests untrusted data from the web and processes it into documentation for downstream AI consumption.
Ingestion points: Research data is gathered from brand websites, trade press, and social platforms like Reddit during Phase 3 and Phase 4 as described in SKILL.md.
Boundary markers: The skill enforces structured markdown headers (e.g., ## What the company sells) but does not use explicit delimiters to isolate untrusted content from instructions.
Capability inventory: The skill has the capability to write files to disk (scripts/lib.py), perform network downloads, and execute Python scripts via the agent.
Sanitization: There is no evidence of sanitization or escaping of the fetched web content before it is written into the research markdown files.

brand-research