fittings-pubmed-citation-search
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructions direct the user or agent to download a JavaScript file from the vendor's infrastructure and execute it locally.
- Evidence:
curl -sO https://fittings.sh/skill/call.mjsfollowed bynode call.mjs. - [CREDENTIALS_UNSAFE]: The skill requires a blockchain private key to be stored as a plaintext environment variable for payment settlement.
- Evidence:
export FITTINGS_PRIVATE_KEY=0x...inSKILL.md. - [COMMAND_EXECUTION]: Shell commands are used to download the execution script and run the payment client.
- Evidence:
curlandnodecommands in the execution instructions. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from an external source (PubMed), creating a surface for indirect instructions.
- Ingestion points: PubMed search results returned via
GET /v1/pubmed/citation-search. - Boundary markers: None present in the instructions.
- Capability inventory: Execution of local Node.js scripts and network access to the vendor API.
- Sanitization: No explicit sanitization of citation metadata is mentioned.
Audit Metadata