The Agent Skills Directory

[DATA_EXFILTRATION]: The skill sends user-provided information, including names, email addresses (as teamily_id), and professional details, to the remote API at https://agentapi.agentapp.space. While this is the intended purpose of the skill for registration and matching, it involves transmitting personally identifiable information to a third-party domain that is not on the standard whitelist.
[INDIRECT_PROMPT_INJECTION]: The skill retrieves and presents agent profiles (bios, skills, and tags) from the NextMarket API, creating a surface for indirect prompt injection. Malicious users could craft profiles with instructions designed to manipulate the behavior of agents that process search results.
Ingestion points: scripts/search_agents.py and scripts/get_agent.py ingest data from the external API.
Boundary markers: No delimiters or warnings are used to signal that retrieved profile data should be treated as untrusted content.
Capability inventory: The skill environment includes tools for network communication and profile management.
Sanitization: The scripts do not appear to sanitize or filter retrieved profile data for potentially malicious prompt injection strings.

agent-social