faion-claude-code
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [EXTERNAL_DOWNLOADS] (LOW): The
mcp-servers/README.mdfile contains a catalog of third-party MCP servers with installation commands usingnpx. Many sources are from trusted organizations like Anthropic, Stripe, and Cloudflare, which limits risk per the trust-scope-rule. Third-party packages are presented for user-initiated manual installation. \n- [COMMAND_EXECUTION] (LOW): Documentation describes how to utilize theBashtool with whitelisting and prefix matching. It also covers the!prefix for commands in slash commands, which is a core platform feature. \n- [PRIVILEGE_ESCALATION] (LOW): Theagents/README.mdfile documents thebypassPermissionsmode. Although it is a high-privilege setting, the documentation explicitly labels it as 'dangerous', providing users with necessary context regarding the security trade-offs. \n- [INDIRECT_PROMPT_INJECTION] (LOW): The documentation defines patterns for interpolating user arguments (e.g.,$ARGUMENTS) directly into commands. \n - Ingestion points:
commands/README.md(via$1,$2,$ARGUMENTS). \n - Boundary markers: Absent in provided templates. \n
- Capability inventory: Full access to
Bash,Task, andWritetools as described in agent configurations. \n - Sanitization: No sanitization logic is present in the provided templates. \n- [CREDENTIALS_UNSAFE] (SAFE): The catalog includes setup instructions for numerous services (Google Ads, Meta, Jira, etc.) but consistently uses placeholders like
...ortokenfor API keys and secrets.
Audit Metadata