Skills
Navigate
skills/openclaw/agent-skills/beam/Gen Agent Trust Hub

beam

Pass

Audited by Gen Agent Trust Hub on Aug 18, 2026

Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTION
Full Analysis
  • [DATA_EXFILTRATION]: The skill's primary function is to read local session transcripts from Claude Code (/.claude/projects/) or Codex (/.codex/sessions/) and upload a redacted version to a remote Gateway endpoint.
  • Evidence: Found in the publish function of scripts/beam and the renderSession function in scripts/beam-session.js.
  • [COMMAND_EXECUTION]: The script executes the cloudflared CLI tool to facilitate authentication and token retrieval for Cloudflare Access-protected endpoints.
  • Evidence: The cloudflaredToken function in scripts/beam uses execFileSync to call cloudflared access token and cloudflared access login.
  • [SAFE]: The skill includes extensive security-focused logic designed to prevent the accidental exposure of sensitive data. It uses a wide array of regular expressions to detect and redact API keys (e.g., OpenAI, Stripe, GitLab, GitHub), authorization headers, private keys, email addresses, and local system paths before any data is sent over the network.
  • Evidence: The redact and sanitizeVisibleText functions in scripts/beam-session.js implement these safeguards.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 18, 2026, 07:50 PM
Security Audit — agent-trust-hub — beam