alby-builder-skill

Pass

Audited by Gen Agent Trust Hub on Jun 25, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill interacts with https://faucet.nwc.dev to provision temporary test wallets and fetch connection secrets. It also references https://esm.sh to load Alby's Bitcoin Connect web components via CDN. These services are vendor-provided or well-known and are central to the skill's intended functionality.
  • [COMMAND_EXECUTION]: A shell script regenerate-types.sh is included for maintaining local typing files, which involves standard file operations (cp, cd) and build tools (yarn). Additionally, the documentation provides curl command examples for manual testing and interaction with the Lightning wallet APIs.
  • [PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection via transaction metadata. Specifically, transaction descriptions received from Nostr relays or invoices (as documented in notifications.md and hold-invoices.md) are ingested into the agent's context. These inputs are used without explicit boundary markers or sanitization, representing a standard risk for applications processing user-provided transaction labels.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 25, 2026, 12:52 PM
Security Audit — agent-trust-hub — alby-builder-skill