The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to capture and process external UI inspiration from sources like Framer and Dribbble. This creates an attack surface where malicious instructions could be embedded in the HTML or metadata of targeted websites. If the agent processes this content (e.g., to generate prompts or summaries), it could be influenced by these hidden instructions.
Ingestion points: Browser-based capture of external landing pages and product sections (SKILL.md).
Boundary markers: None specified in instructions to separate external content from agent instructions.
Capability inventory: The skill has access to shell execution (node, git), file writing, and browser tools.
Sanitization: No explicit sanitization or filtering of captured web content is mentioned.
[COMMAND_EXECUTION]: The skill instructs the agent to run a local script (node scripts/check-ui-inspiration-duplicates.mjs) using a path constructed with a date variable. While the script itself is local, the execution pattern relies on the agent correctly handling the path string.
Evidence: node scripts/check-ui-inspiration-duplicates.mjs articles/YYYY-MM-DD-ui-inspiration-capture/manifest.json in SKILL.md.
[DATA_EXPOSURE]: The skill requires reading potentially sensitive project files such as AGENTS.md and repository status via git status, though this is limited to the current workspace scope.

daily-ui-inspiration-capture