code-architect
Pass
Audited by Gen Agent Trust Hub on Oct 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill provides instructions to execute the
codex-subagentCLI tool. This is the primary functional mechanism of the skill and does not include any unauthorized, privileged, or persistent command patterns. - [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest external context, such as task scopes and issue/PR numbers, which represents an attack surface for indirect prompt injection.
- Ingestion points: Data is passed to the subagent via a shell here-doc in
SKILL.md. - Boundary markers: The skill uses a quoted here-doc (
<<'EOF'), which is a security best practice that prevents the shell from evaluating any variables or commands within the injected text. - Capability inventory: The skill is limited to running the
codex-subagentcommand. - Sanitization: The use of quoted here-docs provides inherent shell-level sanitization against command substitution.
Audit Metadata