The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's supporting scripts (recompose.py, videokit.py, screenrec.py, and arcade-record.py) use the subprocess module to execute external binaries.
These scripts call ffmpeg for video encoding, processing, and rendering walkthroughs.
The arcade-record.py and screenrec.py scripts interact with a local browser-harness CLI to perform browser automation via the Chrome DevTools Protocol (CDP).
These executions are strictly aligned with the skill's primary purpose of capturing and processing media for documentation.
[INDIRECT_PROMPT_INJECTION]: The skill incorporates an attack surface by reading external, untrusted content from GitHub issues to generate documentation.
Ingestion points: Untrusted data enters the agent context through gh issue view commands executed in the enrichment phase (defined in SKILL.md).
Boundary markers: The instructions do not specify explicit boundary markers or delimiters when interpolating issue content into the documentation draft.
Capability inventory: The skill possesses capabilities for file system writes, network access via the gh CLI, and browser automation.
Sanitization: The skill mitigates risks through a 'closed-loop' architecture where a separate, read-only verifier agent checks the generated article against the original acceptance criteria (AC) before any PR is opened.

feature-doc