agent-web-identity
Pass
Audited by Gen Agent Trust Hub on Nov 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes the
icpCLI for authentication and canister interaction. It involves executing shell commands in the background (&) and piping automated input (printf '\n') to handle interactive prompts during the web-linking process. - [EXTERNAL_DOWNLOADS]: The documentation references the official download and installation page for the
icpCLI (https://cli.internetcomputer.org). This is a standard vendor-provided resource for the Internet Computer ecosystem. - [INDIRECT_PROMPT_INJECTION]: The agent is instructed to read and parse the output of commands like
icp identity listand log files from thelinkcommand. This represents a data ingestion surface where external output influences agent logic (e.g., picking unique identity names). The skill mitigates risks by requiring explicit user confirmation before any destructive actions like identity deletion.
Audit Metadata