The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill requires the installation of the 'markdown' Python package via 'pip install markdown'. This is a well-known package from a standard registry.\n- [COMMAND_EXECUTION]: The assemble.py script executes a local preview server (preview-server.py) using os.execvp. This is a legitimate part of the skill's primary functionality to provide a local preview.\n- [PROMPT_INJECTION]: The skill processes .fragment.md files that may contain untrusted data, creating an attack surface for indirect prompt injection.\n
Ingestion points: The scripts/assemble.py script and the defined Edit Workflow read content from .fragment.md files into the agent's context.\n
Boundary markers: There are no explicit boundary markers or instructions in the skill's logic to prevent the agent from interpreting embedded instructions within the fragments as authoritative.\n
Capability inventory: The skill has tool permissions for Read, Write, Edit, and Bash, which allow it to manage files and execute the local preview server.\n
Sanitization: No sanitization or validation of the input Markdown or embedded HTML is performed, allowing for potentially malicious content to be rendered or read back by the agent.

message