The Agent Skills Directory

[COMMAND_EXECUTION]: The skill performs local command execution to validate and release documentation.
It executes npm build and npm check within the web/docs directory to verify the integrity of Astro-based documentation and TypeScript code.
It invokes internal utility scripts (scripts/check-links.sh and scripts/verify-deploy.sh) to perform site-wide link checking and deployment verification.
It manages Git workflows (commit, push, and tagging) for releases, but only when explicit authorization is granted for a specific action, ensuring no unauthorized mutations to the repository occur.
[INDIRECT_PROMPT_INJECTION]: The skill has a defined attack surface as it processes external and user-supplied data to generate documentation.
Ingestion points: Reads free-form user requests in multiple languages (RU/EN) and surveys repository content, including entity READMEs, source code comments, and existing MDX files.
Boundary markers: Implements a strict PLAN — docs block and a structured mode-resolution logic (REVIEW, FIX, PLUGIN, etc.) to ensure the agent remains within the intended scope.
Capability inventory: Possesses file system read/write access for documentation files, shell execution capabilities for build tools and Git, and network access specifically for deployment provenance verification.
Sanitization: Employs a sophisticated multi-agent quorum review process (requiring 2/3 agreement across three independent reviewers) and utilizes fixed MDX templates and repository rules to prevent the adoption of malicious instructions or unsafe code patterns from the analyzed source data.

docs