Subagent Driven Development

Use when executing implementation plans with independent tasks in the current session. Features domain-specialist routing, independent verification gates, an...

This skill is instruction-only and looks coherent for orchestrating per-task subagents. Before enabling it: (1) confirm the referenced prompt files (./implementer-prompt.md, ./verification-prompt.md, etc.) exist and contain safe, intended prompts; (2) verify what 'TodoWrite' and 'superpowers:finishing-a-development-branch' refer to and whether they interact with external services or repos; (3) avoid providing sensitive secrets or workspace state in sessions where the skill will spawn reviewers you don't control; and (4) test in a sandboxed session to ensure the subagent dispatch behavior matches your expectations.

Type: OpenClaw Skill Name: subagent-driven-development-3 Version: 0.1.0 The skill bundle describes a sophisticated subagent-driven software development workflow, emphasizing quality gates, independent verification, and systematic failure recovery. While the core function involves high-risk capabilities like code execution, file system modification, and network interaction (e.g., for API development), these are inherent to software development. The `SKILL.md` instructions are prescriptive and defensive, explicitly outlining 'Red Flags' and 'Never' actions to prevent misuse and enforce secure development practices (e.g., 'Never: Skip reviews', 'Optional: Security/Accessibility review gate'). There is no evidence of intentional harmful behavior, data exfiltration, persistence mechanisms, or prompt injection designed to subvert the agent for malicious purposes. The instructions are clearly aligned with the stated purpose of managing a development process.