The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external files, web links, and repository source code to gather context for documentation, creating a surface for potential indirect prompt injection.
Ingestion points: Stage 1 involves reading user-provided files, external links, and repository artifacts such as implementation code, tests, and schemas.
Boundary markers: The skill includes context isolation for reader testing with subagents but lacks specific boundary markers or instructions to ignore embedded prompts during the primary data ingestion phase.
Capability inventory: The skill utilizes subprocess execution via uv run for documentation build and serve tasks.
Sanitization: No specific sanitization or validation of the ingested external content is mentioned.
[COMMAND_EXECUTION]: The skill integrates documentation preview and validation through the execution of shell commands.
The instructions specify using uv run zensical serve and uv run zensical build within the project environment to maintain documentation quality.

doc-coauthoring