canister-audit
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill clones untrusted Rust repositories from GitHub and executes
cargo buildandcargo testto perform audits and verify findings. Rust compilation allows for arbitrary code execution viabuild.rsscripts and procedural macros, which poses a risk to the execution environment when auditing malicious repositories. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted source code for agentic review. This creates a surface for indirect prompt injection where instructions hidden in the code could influence the AI agent. Evidence chain: 1. Ingestion points: Cloned GitHub repositories; 2. Boundary markers: Absent; 3. Capability inventory: File system access, network access (gh), and command execution (cargo); 4. Sanitization: Absent.
- [DYNAMIC_EXECUTION]: The skill uses
node -eandpython3 -cto perform file operations and write artifacts, bypassing shell restrictions on output redirection defined in the skill's own environment rules. - [COMMAND_EXECUTION]: The workflow relies on executing various system tools including
git,gh,cargo, andsemgrepto process and audit canisters. - [EXTERNAL_DOWNLOADS]: The skill downloads the
semgreppackage from PyPI usingpipand utilizes external binaries like the PocketIC server.
Audit Metadata