implement
Pass
Audited by Gen Agent Trust Hub on Nov 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses sub-agents to perform implementation and testing tasks. These agents are executed via local CLI tools (e.g.,
claude,codex,cursor) as configured inAGENTS_CONFIG.yml. The orchestrator builds these commands using established recipes, ensuring that operations like write access or read-only modes are enforced by the underlying tool when supported. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data such as web pages and external documentation during the investigation phase. It implements a robust 'Trust boundaries' policy that explicitly instructs sub-agents to treat all external content as data rather than instructions, mitigating the risk of indirect prompt injection attacks. It also employs specific sub-agent types (e.g.,
Explorefor read-only tasks) to structurally limit capabilities during research. - [COMMAND_EXECUTION]: The
quality_gates.pyanddecision_form.mjsscripts usesubprocessandspawnto run local utilities likegitand the system browser. These operations are scoped to the local repository and the user's environment, intended for project analysis and interactive configuration, and follow secure practices such as binding to127.0.0.1and using CSRF tokens.
Audit Metadata