The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection. The agent is instructed to fetch and follow documentation, descriptions, and examples from a public marketplace which could be manipulated by third-party app publishers.
Ingestion points: Data retrieved from the /v1/marketplace/apps/{app_id} and /v1/marketplace/search endpoints (specifically description, capability.description, and examples fields).
Boundary markers: None identified; the instructions do not provide delimiters or warnings to ignore instructions embedded in the retrieved marketplace content.
Capability inventory: Wallet registration, signing of SIWE and EIP-712 messages, USDC payment authorizations, and application code deployment (/v1/marketplace/deploy).
Sanitization: No validation or sanitization of the third-party provided documentation strings is described.
[EXTERNAL_DOWNLOADS]: The documentation references the installation of external tools and packages provided by the vendor.
nipper-ai/claude-plugin is referenced via npx skills add to provide persistent API access.
{server}/v1/sdk.tgz is referenced via bun add for developer integration.
[SAFE]: The skill documentation includes proactive safety measures, such as instructing the agent to present Privacy Policy and Terms of Service links for user confirmation and providing clear guidelines on secure private key storage.

nipper