The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill reads and processes documentation files and source code to perform audits and build evaluation projects. This creates a surface where instructions or code examples embedded in the documentation could influence the agent's behavior during the audit or while scaffolding and running the evaluation code.
Ingestion points: Documentation files under docs/src/content/en/ and source files resolved during the audit process (SKILL.md).
Boundary markers: No explicit delimiters or instructions to ignore embedded commands are used when reading documentation content.
Capability inventory: Shell script execution, file system operations, and package installation/execution via pnpm (scripts/eval-setup.sh, scripts/run-checks.sh).
Sanitization: The skill lacks sanitization for documentation content before it is used to create and configure evaluation projects.
[COMMAND_EXECUTION]: The skill executes multiple internal shell scripts that interact with the system and project-specific development tools.
Evidence: The skill invokes scripts like init-run.sh, run-checks.sh, and format-doc.sh which execute CLI tools including pnpm, mkdir, cp, remark, and oxfmt-mdx.
[DYNAMIC_EXECUTION]: The skill dynamically generates configuration files and project structures to test documentation examples in isolated environments.
Evidence: scripts/eval-setup.sh programmatically creates package.json, tsconfig.json, and pnpm-workspace.yaml files at runtime and then executes pnpm install and tsc to verify the generated project.

docs-audit