The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests and analyzes user-controlled web source files (HTML, CSS, JS) to provide design feedback, creating a vulnerability where malicious instructions embedded in the website code could influence the agent's behavior. \n- Ingestion points: Reads and renders local project files as specified in Step 1 of the SKILL.md instructions. \n- Boundary markers: No specific delimiters or safety warnings are implemented to prevent the model from obeying instructions hidden within the analyzed data. \n- Capability inventory: The skill executes local shell commands (git, chrome), manages file system operations, and spawns a local HTTP listener. \n- Sanitization: The look.mjs script includes path normalization and validation to prevent directory traversal attacks during local rendering. \n- [COMMAND_EXECUTION]: The included Node.js script (look.mjs) executes shell commands for 'git' and 'chrome' to perform project discovery and automated browser-based rendering, which is essential for the skill's auditing functionality.

taste