fittings-pubmed-search
Pass
Audited by Gen Agent Trust Hub on Aug 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads an external script (
call.mjs) from the author's domainfittings.sh. This is a vendor-provided resource used to facilitate API calls and payment settlement. - [REMOTE_CODE_EXECUTION]: The instructions direct the agent to execute the downloaded script using
node call.mjs. While the source is the vendor's own domain, executing remote scripts at runtime is a sensitive operation. - [CREDENTIALS_UNSAFE]: The skill requires the user to set a
FITTINGS_PRIVATE_KEYenvironment variable for payment. Although the skill warns about storing keys in plaintext and suggests using a throwaway wallet, this pattern exposes sensitive credentials to the agent's environment. - [INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect injection via the
queryparameter. \n - Ingestion points: The
queryparameter in thepubmed.searchcall inSKILL.md. \n - Boundary markers: Absent. The user input is directly passed as a command-line argument. \n
- Capability inventory: Execution of shell commands via
node call.mjs. \n - Sanitization: No evidence of input sanitization or validation is provided in the instructions.
Audit Metadata