The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to install the markdown Python package via pip for its assembly process.\n- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to the lack of content sanitization.\n
Ingestion points: The assemble.py script reads user-influenced content from .fragment.md and .fragment.html files located in the data/writing/email_drafts/ directory.\n
Boundary markers: No delimiters or boundary instructions are present to prevent the agent or the browser from interpreting malicious instructions or scripts embedded within the fragments.\n
Capability inventory: The skill can launch a local HTTP server (preview-server.py) and execute shell commands to manage the assembly process.\n
Sanitization: Fragment content is directly injected into the shell.html template using string replacement without any escaping or sanitization, making it vulnerable to XSS when viewed in the preview server.\n- [COMMAND_EXECUTION]: The assembly script executes preview-server.py using os.execvp to provide a background preview service for the generated drafts.

message