The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions to download code from an external repository: https://github.com/grp06/openclaw-studio.git. This organization is not recognized as a trusted source and is not clearly linked to the skill author reason-machines.
[REMOTE_CODE_EXECUTION]: The skill recommends using npx -y openclaw-studio@latest, which results in the automatic download and execution of the latest version of the openclaw-studio package from the public npm registry.
[DATA_EXFILTRATION]: The skill accesses sensitive configuration files, specifically ~/.openclaw/openclaw.json and ~/.openclaw/openclaw-studio/settings.json, which are documented to store gateway authentication tokens. Access to these paths represents a data exposure risk.
[DYNAMIC_EXECUTION]: The skill relies on native Node.js modules (better-sqlite3) and provides commands such as npm run verify:native-runtime:repair that trigger the compilation and rebuilding of native binaries on the host system at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill defines a surface for indirect prompt injection through its agent configuration and chat management interfaces.
Ingestion points: Untrusted data enters the agent context via the AgentConfig interface (specifically the systemPrompt and allowedTools fields) in SKILL.md and the sendMessage function which processes arbitrary string content.
Boundary markers: The provided implementation patterns for prompt interpolation do not demonstrate the use of delimiters or instructions to ignore embedded commands within the ingested text.
Capability inventory: The skill environment possesses capabilities for network communication (WebSocket and HTTP fetch), file system operations (via the SQLite RuntimeStore), and local command execution via npm scripts.
Sanitization: While access control is managed via STUDIO_ACCESS_TOKEN middleware, there is no evidence of sanitization, validation, or escaping of the natural language data before it is processed by the AI agents.

openclaw-studio-dashboard