skills/openagentsinc/openagents/l402/Gen Agent Trust Hub

l402

Pass

Audited by Gen Agent Trust Hub on Feb 28, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: Fetches and executes the @lightninglabs/lightning-mcp-server package using the npx utility. Lightning Labs is a well-known and established developer of Lightning Network infrastructure.
  • [COMMAND_EXECUTION]: Spawns numerous shell scripts for node setup and maintenance (e.g., install.sh, start-lnd.sh) located in the local ~/code/lightning-agent-tools directory.
  • [CREDENTIALS_UNSAFE]: References paths to sensitive Lightning Network macaroons (e.g., ~/.lnd/data/chain/bitcoin/mainnet/admin.macaroon) for inspection and security configuration purposes. While these are sensitive paths, their use is central to the skill's purpose of node management.
  • [PROMPT_INJECTION]: The skill interacts with external L402-gated APIs, creating a surface for indirect prompt injection. (1) Ingestion points: Data is received from arbitrary URLs via lnget. (2) Boundary markers: No delimiters are specified to isolate external data from agent instructions. (3) Capability inventory: The agent possesses process:spawn, filesystem:read, and http:outbound capabilities. (4) Sanitization: No sanitization of API responses is documented before the data is incorporated into the agent context.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 28, 2026, 08:50 AM