openclaw-beam
Pass
Audited by Gen Agent Trust Hub on Oct 30, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill is designed to transmit local session transcripts to a remote receiver endpoint. While this involves data transmission, it is the primary intended function and is protected by extensive sanitization logic that redacts secrets, private keys, and local path patterns from the transcript. Payload size limits and truncation are also enforced to prevent excessive data transfer.- [COMMAND_EXECUTION]: The helper script executes the cloudflared CLI tool to obtain authentication tokens for Cloudflare Access. This is a legitimate utility for securing the connection to the receiver gateway and is constrained to specific authentication tasks.- [INDIRECT_PROMPT_INJECTION]: The skill processes session data from external transcript files, creating an ingestion surface for potentially untrusted content.
- Ingestion points: Local JSONL transcript files located in project-specific directories within ~/.claude and ~/.codex.
- Boundary markers: The sanitizeVisibleText function actively identifies and removes internal platform metadata tags like .
- Capability inventory: The skill possesses the capability to perform network POST requests and execute local subprocesses.
- Sanitization: The implementation includes a comprehensive redaction engine that targets diverse sensitive patterns, including JWTs, auth headers, and various cloud provider API keys.
Audit Metadata