The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted metadata from a public token explore feed which could potentially contain malicious instructions intended to influence the agent's behavior during token interaction or launching.
Ingestion points: The GET /agent/launchpad/tokens endpoint (SKILL.md) provides an explore feed containing user-generated descriptions, names, symbols, and social links.
Boundary markers: The skill contains a strong safety contract that explicitly instructs the agent to ignore instructions embedded in untrusted data and requires users to re-type or confirm information.
Capability inventory: The skill possesses significant capabilities including token minting (launch_token), financial swaps (buy, sell), and asset migration (migrate) in SKILL.md.
Sanitization: All write operations require explicit user "yes" confirmation and human-in-the-loop validation of estimated outputs and costs.

openfin-launchpad