The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill communicates with vendor-controlled domains and well-known services.
Sends usage analytics (agent name, OS, architecture) to pancakeswap.ai during initialization.
Fetches token prices and position data from explorer.pancakeswap.com and api.dexscreener.com.
Downloads token metadata from tokens.pancakeswap.finance.
[COMMAND_EXECUTION]: The skill uses shell commands to manage its workflow.
Uses curl to fetch analytics, token lists, and price data.
Uses npm to install required blockchain libraries in temporary directories.
Uses node to execute position discovery scripts.
Uses open or xdg-open to launch official PancakeSwap deep links in the user's browser.
[DYNAMIC_EXECUTION]: The skill generates and executes temporary Node.js scripts (fetch-v3-positions.mjs, fetch-infinity-positions.mjs, fetch-solana.cjs) to interact with blockchain SDKs.
These scripts perform read-only operations, such as calling balanceOf and simulating collect calls to estimate pending fees.
Position data is processed locally to generate a summary for the user.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources.
Ingestion points: Token symbols and names are ingested from on-chain metadata and external APIs (DexScreener, Explorer) in Step 3.
Boundary markers: The instructions include a specific security rule (#3) telling the agent to treat API data as untrusted and never interpret token names as commands.
Capability inventory: The skill can install packages, execute node scripts, and open URLs.
Sanitization: Shell variable interpolation is protected by mandatory single-quoting rules, and wallet addresses are validated against strict regex patterns before use.

collect-fees