Subagent Driven Development
This is a structured workflow for executing implementation plans by dispatching a fresh subagent per task, then running two mandatory review passes: spec compliance first, code quality second. The controller stays in the current session and coordinates everything through careful context packets instead of dumping full chat history. Each implementer reports status (DONE, BLOCKED, etc.) and you handle it appropriately before moving on. The discipline here is in the two-stage review loop and never skipping re-reviews when issues are found. It's opinionated about sequencing and includes templates for the prompts. Good for plans with independent tasks where you want tight control and don't mind the overhead of multiple review cycles per task.
- GitHub stars
- 492
- last updated
- Jun 11, 2026
Install
$ npx -y skills add ganyuanran/aegis --skill subagent-driven-development --agent claude-codeInstalls into .claude/skills of the current project.
Files
Execute
→ Have an implementation plan with independent tasks? → Fresh subagent per task + two-stage review.
- Read plan, extract all tasks, create TodoWrite
- Per task: dispatch implementer → answer questions → implementer completes
- Review stage 1 (spec compliance) → fix gaps → re-review until ✅
- Review stage 2 (code quality) → fix issues → re-review until ✅
- Coordinator verifies, commits the coherent task, updates checkpoint/drift → next task → All tasks done: final review → verification receipt; branch finishing only when needed.
Subagent-Driven Development
Execute plan by dispatching fresh subagent per task, with two-stage review after each: spec compliance review first, then code quality review.
Why subagents: You delegate tasks to specialized agents with isolated context. By precisely crafting their instructions and context, you ensure they stay focused and succeed at their task. They should never inherit your session's context or history — you construct exactly what they need. This also preserves your own context for coordination work.
Sponsors
Agent CaffeineSponsoredClose your lid, come back to finished work
SnugSponsoredThe clipboard manager for the AI eraPromote your product hereReach 125k+ visitors monthly
WaxSponsoredBlock anything with supervision, without erasing your iPhonePromote your product hereReach 125k+ visitors monthly
knobsSponsoredSee your site on any phone, any day, in any languageDetails
- Repository
- github.com/ganyuanran/aegis
- Categories
- AI & Agent Building, Automation & Workflows
- License
- MIT
- First seen
- Jun 11, 2026
More from ganyuanran/aegis
All 22- Systematic Debugging
- Test Driven Development
- Update Aegis
- Using Aegis
- Using Git Worktrees
- Verification Before Completion
- Writing Plans
- Writing Skills
- Anti Entropy Governance
- Brainstorming
- Communicating Concisely
- Dispatching Parallel Agents
- Establishing Project Context
- Executing Plans
- Finishing A Development Branch
- First Principles Review
- Goal Framing
- Long Task Continuation
- Receiving Code Review
- Recording Architecture Decisions
- Requesting Code Review
Recommended
More AI & Agent Building
ai-daily-newsgeekjourneyx/ai-daily-skillFetches AI news from smol.ai RSS and generates structured markdown with intelligent summarization and categorization. Optionally creates beautiful HTML webpages0skills.sh installs
subagent-driven-analysisk-dense-ai/science-superpowersWhen you have a frozen, pre-registered analysis plan with independent steps, this skill runs them one by one using fresh subagents per...0skills.sh installs
add-new-entrykimtth/awesome-azure-openai-llmWorkflow and tools for adding new entries from temp.md to the section files. Includes legend format, section reference, code tools, and common pitfalls. USE FOR0skills.sh installsclassify-temp-entries-to-sectionkimtth/awesome-azure-openai-llmClassification guidelines for entries in temp_entries.md. Each entry have its own title with the markdown file name and section name in temp_entries.md. USE FOR0skills.sh installsupdate-app-countkimtth/awesome-azure-openai-llmWorkflow for updating the popular LLM applications pool (section/x_llm_apps.md) using get_app_list_by_github_star.py. Covers full refresh, alternate exports, to0skills.sh installs
forgecad-prepare-promptkostard/forgecad-public-kitTurn a fuzzy physical product, mechanism, or CAD artifact request into a concrete manufacture-realistic prototype ForgeCAD build brief and a single master promp0skills.sh installs