The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The API documentation specifies that the token/launch and token/launch/batch endpoints require a Solana private_key (formatted as a JSON array, base64, or base58 string) to be included in the request body for token minting operations. While this is a functional requirement for the service, it necessitates the agent handling and transmitting highly sensitive private wallet credentials.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation and update of marketplace items by ingesting and transmitting arbitrary text and code provided by the user.
Ingestion points: The agent (code), prompt (instructions), description, and useCases fields within the POST /api/add-agent and POST /api/add-prompt endpoints.
Boundary markers: The documentation does not specify the use of delimiters or warnings to prevent the AI from executing instructions that may be embedded within the user-supplied agent code or prompts.
Capability inventory: The skill demonstrates capabilities for network writes (POST requests) to the swarms.world domain via curl examples.
Sanitization: The reference does not describe sanitization, filtering, or validation processes to be applied to the agent or prompt content prior to API transmission.

swarms-marketplace-api