The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions rely heavily on the execution of the agent CLI tool to perform operations.
It utilizes agent x402-pay to make network requests that trigger financial transactions (spending USDC from the user's wallet).
It provides instructions for managing local sensitive assets through commands like agent wallet login, agent wallet list, and agent balances.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from various external providers, creating an attack surface for indirect prompt injection.
Ingestion points: Untrusted content from the web is brought into the agent context through endpoints such as exa/search, searchapi/google, news/headlines, and firecrawl/scrape (found in SKILL.md).
Boundary markers: There are no explicit instructions or delimiters provided to guide the agent in distinguishing between operational instructions and the data fetched from these external sources.
Capability inventory: The agent has the capability to execute shell commands (agent x402-pay), perform network operations, and interact with the user's financial wallet.
Sanitization: The skill does not mention or implement any sanitization or validation of the data retrieved from the proxied search and scraping services.

polygon-discovery