acp-checkout-mcp
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to fetch documentation, release information, and specification files from external sources.
- It fetches release data and OpenRPC JSON specifications from
github.com/agentic-commerce-protocol. - It fetches checkout operation semantics from
developers.openai.com. - It fetches protocol specifications from
modelcontextprotocol.io. - [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection due to its reliance on external, potentially untrusted content to drive its primary functions.
- Ingestion points: The agent is instructed to fetch and parse external OpenRPC specifications (
openrpc.agentic_checkout.json) and SDK documentation from GitHub and official websites (specified inSKILL.md). - Boundary markers: None are present; there are no instructions to the agent to treat the fetched specification data as data only or to ignore embedded instructions.
- Capability inventory: The skill allows access to high-privilege tools including
Bash,Write,Edit, andWebFetchinSKILL.md. These tools are intended to be used to implement server code based on the fetched data. - Sanitization: There is no mention of sanitizing or validating the fetched JSON-RPC descriptions or documentation before the agent uses them to generate tool schemas or server logic.
Audit Metadata