The Agent Skills Directory

[PROMPT_INJECTION]: The skill processes untrusted repository files containing instructions (e.g., AGENTS.md, CONTRIBUTING.md, and .cursorrules), which creates a surface for indirect prompt injection. \n
Ingestion points: The skill reads various configuration and instruction files using Read, Glob, and Grep as described in references/agent-and-contributing.md and agents/inventory-agent.md. \n
Boundary markers: The skill does not explicitly mandate the use of delimiters or 'ignore instructions' warnings when reading these repository-provided files to protect its own task context. \n
Capability inventory: The skill utilizes tools to read and modify repository content and can execute local project scripts using pnpm and ripgrep. \n
Sanitization: No explicit content sanitization or validation of the ingested file content is mentioned in the playbooks. \n- [EXTERNAL_DOWNLOADS]: The skill includes references to various external documentation sources and industry blogs to establish its operating principles. \n
Evidence: Links to resources from trusted organizations such as GitHub, OpenAI, and Vercel are used to define documentation best practices. \n
Reasoning: These references are from well-known technology companies and official repositories, which is standard practice for high-quality documentation skills. \n- [COMMAND_EXECUTION]: The skill utilizes standard local development tools for repository discovery and documentation validation. \n
Evidence: Usage of rg (ripgrep) for file discovery and pnpm for running documentation checks (e.g., pnpm docs:check-links) is documented in references/agent-and-contributing.md and references/openclaw.md. \n
Reasoning: These commands are routine operations within a development environment for documentation maintenance and do not involve remote code execution or suspicious privileges.

technical-documentation