fable-codex-setup
Pass
Audited by Gen Agent Trust Hub on Sep 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs the user to install the
@openai/codexCLI tool and associated plugins. These are resources associated with the official OpenAI organization. - [COMMAND_EXECUTION]: The
codex-executorsub-agent is designed to execute thecodex execcommand via theBashtool to perform bulk code implementation tasks as part of its core functionality. - [INDIRECT_PROMPT_INJECTION]: The orchestration architecture creates an attack surface where untrusted user input is processed by the primary agent and delegated to sub-agents.
- Ingestion points: User-defined goals and context files are processed by the Fable 5 orchestrator (SKILL.md, templates/CLAUDE-orchestration.md).
- Boundary markers: Not explicitly defined in the provided sub-agent templates for input separation.
- Capability inventory: The
codex-executorsub-agent has access to theBashtool to modify the filesystem and execute CLI commands (templates/agents/codex-executor.md). - Sanitization: The skill mitigates risks by instructing the agent to always verify sub-agent output using
git diff --statand automated success-criterion commands before integration.
Audit Metadata