agentpmt-docs-and-content

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill (“AgentPMT Platform Search”, actions search/global_search/recent) ingests AgentPMT’s public content at runtime based on user-supplied query and returns its items (with url/snippets) for the agent to use, so an outsider can influence which free text content is pulled in from an externally authored public surface (e.g., community/public pages hosted on AgentPMT).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). This skill instructs agents to call the AgentPMT MCP/REST endpoints (https://api.agentpmt.com/mcp/ and https://api.agentpmt.com/products/purchase) at runtime to fetch live schema and get_instructions which are then treated as authoritative guidance and can directly control agent prompts/behavior.