The Agent Skills Directory

[COMMAND_EXECUTION]: The skill spawns a subprocess using Bun.spawn to invoke the signing/signing.ts utility for BIP-322 message signing. The command structure and the message format are controlled internally by the script, which mitigates standard shell injection risks.
[INDIRECT_PROMPT_INJECTION]: The file-signal command accepts user-provided text for news headlines (up to 120 characters) and content (up to 1000 characters). This data is processed and sent to the aibtc.news API. This represents a vulnerability surface where malicious instructions could be embedded in signals to target the platform or other agents consuming the news feed.
Ingestion points: User-supplied --headline and --content arguments in aibtc-news.ts (file-signal command).
Boundary markers: The skill enforces character length limits but lacks explicit delimiters or instructions for downstream consumers to ignore embedded commands.
Capability inventory: The skill has the capability to execute subprocesses (signing), perform network requests (API interaction), and access wallet private keys for transaction signing.
Sanitization: Basic validation is performed on input lengths and JSON structures for sources and tags, but no content filtering for natural language instructions is present.
[CREDENTIALS_SAFE]: The skill interacts with the wallet's private key to sign Stacks transactions for sBTC payments (x402 protocol) and uses the local signing skill for BIP-322 authentication. These sensitive operations are handled in memory using established libraries (@stacks/transactions) and trusted local skill components.

aibtc-news