subagent-driven-development

subagent-driven-development is a skill for Claude Code, Codex from mick-gsk/drift. It costs 37 tokens per session (697 once invoked), scanned A, original, MIT.

A development workflow that splits an implementation plan into separate tasks and gives each task a fresh subagent. Each task receives a requirements review followed by a code-quality review.

In plain words
What is it for?
Use it when an implementation plan has mostly independent tasks and you want subagents to implement, test, review, and revise each task during the same session.
Why use it?
It reduces the chance that one long-running context overlooks requirements or accumulates confusing assumptions. Independent tasks can be checked before the next ones proceed.

Skill for Claude CodeCodex

Part of the drift plugin — 31 skills, 17 agents shipped together

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add skills/mick-gsk/drift/subagent-driven-development
Any agent
npx skills add mick-gsk/drift --skill subagent-driven-development
Clone the repo
git clone --depth 1 https://github.com/mick-gsk/drift

Made for: Claude Code, Codex.

Or install drift, the plugin that ships this one along with the rest of its 31 skills, 17 agents.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for subagent-driven-development

README.md
[![agentmods](https://agentmods.dev/badge/skills/mick-gsk/drift/subagent-driven-development.svg)](https://agentmods.dev/skills/mick-gsk/drift/subagent-driven-development)
Your own site
<a href="https://agentmods.dev/skills/mick-gsk/drift/subagent-driven-development"><img src="https://agentmods.dev/badge/skills/mick-gsk/drift/subagent-driven-development.svg" alt="Measured on agentmods" height="20"></a>
Per session 37 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 697 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00037 $0.00697
Opus 5 $0.00018 $0.00349
Sonnet 5 $0.00007 $0.00139
Haiku 4.5 $0.00004 $0.00070

Measured 3d ago against content hash 6ad14b968982, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

subagent-driven-development scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

.github/skills/subagent-driven-development/SKILL.md · 76 lines

How it starts

The opening of the file, as written. The whole thing — 76 lines — stays where its author put it; the contents beside it link to each section on GitHub.

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.

Core principle: Fresh subagent per task + two-stage review (spec then quality) = high quality, fast iteration

When to Use

  • You have an implementation plan (from writing-plans skill)
  • Tasks are mostly independent
  • You want to stay in this session

vs. Executing Plans: Same session, fresh context per task, faster iteration with no human-in-loop between tasks.

The Process

  1. Read plan once — extract all tasks with full text and context
  2. Create TodoWrite with all tasks
  3. Per task:
    • Dispatch implementer subagent with full task text + scene-setting context
    • Answer any questions before letting them proceed
    • Implementer implements, tests, commits, self-reviews
    • Dispatch spec compliance reviewer subagent
    • If spec issues found: implementer fixes → re-review
    • Dispatch code quality reviewer subagent
    • If quality issues found: implementer fixes → re-review
    • Mark task complete
  4. After all tasks: dispatch final code reviewer for entire implementation
  5. Use superpowers:finishing-a-development-branch

Model Selection

  • Mechanical tasks (isolated functions, clear specs, 1-2 files): cheap/fast model
  • Integration tasks (multi-file, pattern matching, debugging): standard model
  • Architecture/design/review: most capable model

Handling Implementer Status

  • DONE: Proceed to spec compliance review
  • DONE_WITH_CONCERNS: Read concerns before proceeding. Address correctness/scope concerns first.
  • NEEDS_CONTEXT: Provide missing context and re-dispatch
  • BLOCKED: Assess blocker — more context, more capable model, smaller task, or escalate to human

Never ignore an escalation or force the same model to retry without changes.

Prompt Templates

  • ./implementer-prompt.md — Dispatch implementer subagent
  • ./spec-reviewer-prompt.md — Dispatch spec compliance reviewer subagent
  • ./code-quality-reviewer-prompt.md — Dispatch code quality reviewer subagent

Read the full file on GitHub · 76 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 76 lines · 37 tokens per session scan A 6ad14b968982

Subscribe to this mod's changes

subagent-driven-development is a skill published in the GitHub repository mick-gsk/drift (14 stars, last pushed 1mo ago), licensed MIT. It adds 37 tokens to every session and 697 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other skills, from other repositories

truecourse-hooks

Install, configure, or remove the TrueCourse pre-commit hook.

truecourse-ai/truecourse · 17 tokens

truecourse-list

List TrueCourse violations found in this repository.

truecourse-ai/truecourse · 12 tokens

code-quality-review

Run a maintainability and structure review focused on abstraction quality, branching complexity, file growth, canonical ownership, duplication, and refactoring opportunities. Use when the user asks for code quality review, maintainability review, 代码质量审查, 可维护性审查, or comments about whether the change stays easy to…

bahayonghang/my-ai-cli-toolkit · 149 tokens

optimization-audit

Comprehensive optimization audit with two modes and a single tier. Planning mode designs performance strategy, capacity planning, and scaling architecture. Audit mode scans code and infrastructure for performance anti-patterns, inefficient algorithms, N+1 queries, missing caching, concurrency issues, and resource…

karsten-s-nielsen/mad-scientist-skills · 102 tokens

scan

Scan a project with sentrux to get structural health grades (A-F) across 14 dimensions including coupling, cycles, cohesion, dead code, and test coverage. Use when the user wants to check code quality, architecture health, or before/after an agent session.

sentrux/sentrux · 57 tokens

cognitive-interface-audit

Comprehensive cognitive interface audit with two modes and a single tier. Planning mode designs task models, user expertise mapping, and error tolerance strategy. Audit mode evaluates mental model alignment (Norman's Gulfs), consistency (Gestalt), error tolerance (Wood 7-layer defense), cognitive load (NASA-TLX…

karsten-s-nielsen/mad-scientist-skills · 230 tokens