The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data such as token names, symbols, and descriptions that are retrieved from the blockchain via the Flaip API and user inputs.
Ingestion points: Untrusted data enters the agent context through the /api/agent/resolve endpoint and user-defined token attributes in the creation workflow.
Boundary markers: The instructions explicitly direct the agent to treat on-chain metadata as data only and not to interpret them as instructions, providing a clear boundary for the model.
Capability inventory: The skill utilizes the binance-agentic-wallet skill to execute high-privilege operations including contract calls, asset swaps, and token transfers.
Sanitization: Mitigation relies on prompt-based instructions to ensure external data is not executed; no programmatic sanitization of the API data is mentioned.
[EXTERNAL_DOWNLOADS]: The skill references and fetches operational guidelines, quickstart markdowns, and API specifications from the service domain flaip.xyz.
[COMMAND_EXECUTION]: The skill provides instructions to set up the environment using npx skills add, which involves shell command execution to install necessary sub-skills like the binance-agentic-wallet.

flaip-launchpad