The Agent Skills Directory

[COMMAND_EXECUTION]: The skill requires the agent to run local Python scripts (scripts/render-response-html.py and scripts/render-intake-form-html.py) to generate HTML documents. It also utilizes the rg (ripgrep) utility to search through cached files.\n- [DATA_EXFILTRATION]: The skill instructs the agent to read and write files within ~/Documents/Codex/.chrisai/cache/. This access to the user's home directory outside the standard project root introduces a risk of exposure or modification of sensitive user data.\n- [DYNAMIC_EXECUTION]: The render-intake-form-html.py script generates HTML artifacts that contain embedded JavaScript for form logic. Although labels are escaped, the creation of files with active scripts on the local system creates a client-side execution surface.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted chat history and form responses to build persistent artifacts.\n
Ingestion points: Conversation transcripts in workflows/chat-session-caching-and-recall.md and pasted form text in workflows/html-form-intake.md.\n
Boundary markers: The workflows use structured templates but do not include explicit instructions for the agent to ignore malicious prompts embedded in the ingested data.\n
Capability inventory: File system read/write access and execution of rendering scripts.\n
Sanitization: Rendering scripts employ html.escape and link validation to sanitize data before it is rendered into HTML.

chrisai-chatting