The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the user to clone a repository from https://github.com/xmanrui/OpenClaw-bot-review.git. This source is not among the verified or trusted organizations, introducing a risk of executing unvetted code.
[COMMAND_EXECUTION]: The installation process requires executing several shell commands including git clone, npm install, npm run start, and docker run. Running installation scripts (npm install) on a third-party codebase allows for arbitrary code execution on the host machine.
[CREDENTIALS_UNSAFE]: The dashboard is designed to read and process ~/.openclaw/openclaw.json, which explicitly stores sensitive information such as app_secret for Feishu and api_key for model providers like OpenAI. The provided code snippets show these secrets being read into memory and used for platform connectivity tests.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection as it ingests untrusted data from session files (~/.openclaw/sessions/) and displays it in a web dashboard.
Ingestion points: Reads all JSON files in ~/.openclaw/sessions/{agent-name}/ via getAgentSessions in lib/sessions.ts.
Boundary markers: None identified in the provided code snippets or instructions.
Capability inventory: Performs network requests via fetch (e.g., in Gateway Health Check and testPlatformConnection) and reads local files using the fs module.
Sanitization: There is no evidence of sanitization or escaping of the session content before it is parsed and rendered in the React/Next.js frontend.

openclaw-bot-review-dashboard