The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes instructions to download and execute an installation script for the uv tool.
Evidence: curl -LsSf https://astral.sh/uv/install.sh | sh in SKILL.md.
Context: The script is hosted on astral.sh, the official domain for the well-known uv Python utility.- [EXTERNAL_DOWNLOADS]: The skill interacts with the mermaid.ink API to fetch image renders of diagrams.
Evidence: Step 3c describes using urllib.request to fetch data from https://mermaid.ink/img/.- [COMMAND_EXECUTION]: The skill executes system commands such as which, python3, and uvx to manage the environment and run the conversion tool.
Evidence: Bash blocks in Step 2 and Step 4.- [DATA_EXFILTRATION]: The skill reads local Markdown files and transmits their contents to the Feishu platform.
Evidence: The primary functionality involves reading a <MARKDOWN_FILE_PATH> and using feishu-docx to create a document.- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it parses user-provided Markdown files to determine the document title and render diagrams.
Ingestion points: Untrusted Markdown files provided by the user.
Boundary markers: Absent; the agent reads the file content directly to extract the first heading.
Capability inventory: File system access and command execution through the feishu-docx and uvx tools.
Sanitization: No sanitization is applied to the Markdown content before it influences the agent's logic or is sent to external APIs.

md-to-feishu