god-tibo-imagen

Warn

Audited by Gen Agent Trust Hub on Apr 24, 2026

Risk Level: MEDIUMCREDENTIALS_UNSAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
  • [CREDENTIALS_UNSAFE]: The skill is designed to read and reuse authentication tokens stored in the local file ~/.codex/auth.json. This file contains sensitive session data for the Codex/ChatGPT service.
  • [EXTERNAL_DOWNLOADS]: The skill requires downloading and installing third-party packages from the NPM and PyPI registries (god-tibo-imagen). It also suggests adding skills via a remote GitHub repository.
  • [COMMAND_EXECUTION]: The skill relies on shell command execution for its primary functionality, including package management (npm, pip, npx) and the core image generation tool (gti).
  • [DATA_EXFILTRATION]: By reading local credential files while having permissions for WebFetch and Bash, the skill possesses the technical capability to exfiltrate authentication tokens to external servers.
  • [PROMPT_INJECTION]: The skill processes untrusted inputs from users and external image files, creating a surface for indirect prompt injection.
  • Ingestion points: CLI prompt strings and reference images passed via the --input argument.
  • Boundary markers: The skill lacks explicit boundary markers or instructions to disregard embedded commands in user-provided content.
  • Capability inventory: Access to Bash, WebFetch, Read, and Write tools across its lifecycle.
  • Sanitization: No mention of input validation or sanitization of the multimodal data processed at runtime.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Apr 24, 2026, 02:47 AM