chofex-hackathon

Pass

Audited by Gen Agent Trust Hub on Oct 22, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill relies on the andes CLI tool to perform critical operations including account authentication, application registration, and challenge management. These commands are executed locally to interact with the hackathon's backend services.
  • [EXTERNAL_DOWNLOADS]: The agent is instructed to proactively install or upgrade the hacktheandes-cli package from the NPM registry and use bun install for setting up specific technical challenge dependencies. These downloads are tied to the skill's primary purpose and originate from standard package registries.
  • [DYNAMIC_EXECUTION]: As part of the event's admission process, the agent is designed to write, test, and evaluate JavaScript and TypeScript code locally. This dynamic code generation is a central feature for assisting participants with algorithmic challenges.
  • [INDIRECT_PROMPT_INJECTION]: The skill processes data from CLI outputs and challenge materials, creating an injection surface. Ingestion points: Untrusted data enters via CLI outputs for status, schemas, and challenge notebooks. Boundary markers: The skill mandates the use of --output json and versioned envelopes to provide structure. Capability inventory: The agent has the ability to execute shell commands, write to the filesystem (using restricted mode-600 temporary files), and run generated scripts. Sanitization: The agent follows instructions to normalize URLs and handles and to strictly map JSON keys from external responses rather than interpreting raw text directly.
Audit Metadata
Risk Level
SAFE
Analyzed
Oct 22, 2026, 09:22 PM