tig-innovator
Warn
Audited by Gen Agent Trust Hub on Mar 4, 2026
Risk Level: MEDIUMCREDENTIALS_UNSAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: The skill requires a
TIG_PRIVATE_KEYenvironment variable for mainnet submissions. Handling raw private keys in environment variables poses a risk of accidental exposure or logging of highly sensitive credentials.- [COMMAND_EXECUTION]: Thetestcommand executes code, which may be modified or generated by an LLM, using Docker benchmarks via theBashtool. While Docker provides isolation, the execution of non-static code remains a significant security surface.- [EXTERNAL_DOWNLOADS]: The skill downloads external algorithm source code from the TIG network API (api.tig.foundation) for the purpose of analysis and optimization.- [PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted code from external sources to be analyzed by an LLM. This code could contain instructions designed to manipulate the optimization process or the agent's actions. - Ingestion points: External algorithm source code downloaded via
tig-innovator analyze. - Boundary markers: None identified in the skill documentation to separate instructions from the data being analyzed.
- Capability inventory: Includes file system access (
Writetool), shell execution (Bashtool), and network/blockchain submission capabilities. - Sanitization: No mentions of sanitization or safety validation of the downloaded code before processing by the LLM.
Audit Metadata