doubt-driven-development
Pass
Audited by Gen Agent Trust Hub on Aug 9, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill facilitates the execution of external CLI tools (gemini, codex) via shell commands for cross-model verification. * Evidence: Shell commands 'which gemini', 'gemini --version', and piping artifacts to 'codex' or 'gemini' in 'SKILL.md'. * Context: The skill mandates explicit user authorization before execution and requires defensive measures like piping input from temporary files to prevent shell command injection from potentially malicious code artifacts.
- [PROMPT_INJECTION]: The skill processes untrusted artifacts and contracts from external sources, creating an indirect prompt injection surface. * Ingestion points: 'ARTIFACT' and 'CONTRACT' data inputs in 'SKILL.md'. * Boundary markers: The skill instructs the agent to place untrusted content within specific adversarial prompt delimiters. * Capability inventory: Use of subprocess calls to external CLI tools (gemini, codex) and subagent invocations. * Sanitization: The skill mandates the use of read-only sandboxes for external tools and explicitly warns against shell interpolation to prevent command injection from malicious artifact content.
Audit Metadata