The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from untrusted external sources, creating a potential surface for indirect prompt injection attacks.
Ingestion points: Untrusted data enters the agent context through BOLT-11 invoices, LNURL strings, and lightning addresses processed by components like bc-pay-button (references/bitcoin-connect/bitcoin-connect.md) and functions like requestInvoice (references/lightning-tools/lnurl.md).
Boundary markers: The provided code templates do not include explicit delimiters or instructions to ignore embedded commands within user-supplied payment strings.
Capability inventory: The skill grants the agent significant capabilities, including network access via Nostr relays, payment settlement via the NWCClient (references/nwc-client/nwc-client.md), and automated HTTP 402 challenge handling via fetch402 (references/lightning-tools/402.md).
Sanitization: While the skill validates cryptographic formats (e.g., payment hashes), it does not sanitize descriptive text or success messages which could be manipulated to influence agent behavior.
[EXTERNAL_DOWNLOADS]: The skill documentation references and utilizes external code and services.
Evidence: The skill instructs users to install NPM packages from the @getalby scope and suggests fetching web components from the well-known esm.sh CDN (references/bitcoin-connect/bitcoin-connect.md).
Evidence: Automated testing scripts reference the faucet.nwc.dev service to programmatically generate test wallets (references/automated-testing.md). These are established resources within the vendor's ecosystem.

alby-bitcoin-builder