The Agent Skills Directory

[COMMAND_EXECUTION]: The script quip-to-markdown.py invokes system utilities curl, ffmpeg, and file via subprocess.run. The implementation includes input validation (regex checks for IDs) and path sanitization to mitigate command injection risks. The push-to-wiki.sh script executes git clone and git push using provided repository URLs.\n- [INDIRECT_PROMPT_INJECTION]: The skill ingests HTML content from the Quip API and processes it into Markdown. While this represents a surface for indirect prompt injection, the skill employs sanitization through the html2text library and strict path validation (sanitize_path) for all downloaded assets. Ingestion points: urllib.request.urlopen in quip-to-markdown.py. Boundary markers: Absent. Capability inventory: subprocess.run (ffmpeg, curl, file), file writing, and git. Sanitization: sanitize_path function and re.match validation.\n- [EXTERNAL_DOWNLOADS]: The documentation in references/gitlab-mcp-config.md references the external Node.js package @zereight/mcp-gitlab to be run via npx.\n- [CREDENTIALS_UNSAFE]: The documentation provides guidance on standard SSH configuration files (~/.ssh/config and ~/.ssh/id_ed25519). These are documented for the user to configure their own environment and are not accessed by the skill's scripts. The scripts correctly read the QUIP_TOKEN from an environment variable and pass it to subprocesses via stdin to avoid exposure in the process list.

quip-to-gitlab-wiki