geo-forward-geocode
Pass
Audited by Gen Agent Trust Hub on Aug 18, 2026
Risk Level: SAFECREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONREMOTE_CODE_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructions suggest downloading an executable script from 'https://fittings.sh/skill/call.mjs'. While this is the vendor's own domain, downloading and running remote code is a security consideration.
- [CREDENTIALS_UNSAFE]: The skill uses the FITTINGS_PRIVATE_KEY environment variable to manage a blockchain private key. Storing sensitive keys in environment variables is common but carries risk if the environment is shared or compromised.
- [COMMAND_EXECUTION]: The skill requires running 'node call.mjs', granting the script the ability to perform network requests, file system reads, and cryptographic operations on the host.
- [REMOTE_CODE_EXECUTION]: The 'call.mjs' script uses dynamic 'import()' to load external libraries at runtime and verifies its own integrity by fetching a hash from the vendor's server.
- [PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection. Ingestion points: address argument in call.mjs. Boundary markers: Not present. Capability inventory: Includes network access (fetch), file system access (readFileSync), and dynamic code loading. Sanitization: Inputs are URL-encoded for delivery, but standardized outputs returned from the API are processed by the agent without further verification.
Audit Metadata