subagent-driven-development

# Subagent-Driven Development

Execute plan by dispatching a fresh implementer subagent per task, a task review (spec compliance + code quality) after each, and a broad whole-branch review at the end.

**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.

**Core principle:** Fresh subagent per task + task review (spec + quality) + broad final review = high quality, fast iteration

**Narration:** between tool calls, narrate at most one short line — the
ledger and the tool results carry the record.

**Continuous execution:** Do not pause to check in with your human partner between tasks. Execute all tasks from the plan without stopping. The only reasons to stop are the four named below, or all tasks complete. "Should I continue?" prompts and progress summaries waste their time — they asked you to execute the plan, so execute it.

**Rulings, not stalls.** A running plan does not wait on a human. Conflicts,
ambiguities, plan defects, a cap you would have asked to exceed — decide
them. The spec is the binding authority, the plan is its argument, and your
judgment settles what neither answers. Record every decision in the ledger as
`Ruling: <what you decided> — <why> — <what it costs if wrong>`, and keep
going. A wrong ruling costs rework your human partner can see and undo; a
session parked on a question costs their whole day and buys nothing.

Four things stop you, and only these: an irreversible or destructive
operation; a security-sensitive action; a side effect outside this worktree
that norms say you ask about first (a merge, a push to a shared branch, a
publish); and a plan so broken that every path forward is a guess. For those,
stop and ask.

## When to Use

```dot
digraph when_to_use {
    "Have implementation plan?" [shape=diamond];
    "Tasks mostly independent?" [shape=diamond];
    "Partner chose inline, or no subagent tool?" [shape=diamond];
    "subagent-driven-development" [shape=box];
    "executing-plans" [shape=box];
    "Manual execution or brainstorm first" [shape=box];

    "Have implementation plan?" -> "Tasks mostly independent?" [label="yes"];
    "Have implementation plan?" -> "Manual execution or brainstorm first" [label="no"];
    "Tasks mostly independent?" -> "Partner chose inline, or no subagent tool?" [label="yes"];
    "Tasks mostly independent?" -> "Manual execution or brainstorm first" [label="no - tightly coupled"];
    "Partner chose inline, or no subagent tool?" -> "executing-plans" [label="yes"];
    "Partner chose inline, or no subagent tool?" -> "subagent-driven-development" [label="no"];
}
```

**vs. Executing Plans (inline):**
- Fresh subagent per task (no context pollution) instead of one context doing every task
- Review after each task (spec compliance + code quality) instead of only at the end
- Costs a fresh context per task and per review; inline costs one context plus one final reviewer
- Both run in this session, share the same plan workspace and ledger, and never pause between tasks

## The Process

```dot
digraph process {
    rankdir=TB;

    subgraph cluster_per_task {
        label="Per Task";
        "Dispatch implementer subagent (./implementer-prompt.md)" [shape=box];
        "Implementer asks questions?" [shape=diamond];
        "Answer questions, provide context" [shape=box];
        "Implementer implements, tests, commits, self-reviews" [shape=box];
        "Generate review package, dispatch task reviewer (./task-reviewer-prompt.md)" [shape=box];
        "Spec ✅ and quality approved?" [shape=diamond];
        "Fix loop: read references/fix-loop.md before every round" [shape=box];
        "Append completion to ledger, mark todo complete" [shape=box];
    }

    "Setup: worktree, ledger check, read plan, pre-flight review" [shape=box];
    "More tasks remain?" [shape=diamond];
    "Final review and finish: read references/final-review.md" [shape=box];
    "Final review clean: delete this plan's workspace" [shape=box];
    "Use superpowers:finishing-a-development-branch" [shape=box style=filled fillcolor=lightgreen];

    "Setup: worktree, ledger check, read plan, pre-flight review" -> "Dispatch implementer subagent (./implementer-prompt.md)";
    "Dispatch implementer subagent (./implementer-prompt.md)" -> "Implementer asks questions?";
    "Implementer asks questions?" -> "Answer questions, provide context" [label="yes"];
    "Answer questions, provide context" -> "Implementer implements, tests, commits, self-reviews";
    "Implementer asks questions?" -> "Implementer implements, tests, commits, self-reviews" [label="no"];
    "Implementer implements, tests, commits, self-reviews" -> "Generate review package, dispatch task reviewer (./task-reviewer-prompt.md)";
    "Generate review package, dispatch task reviewer (./task-reviewer-prompt.md)" -> "Spec ✅ and quality approved?";
    "Spec ✅ and quality approved?" -> "Append completion to ledger, mark todo complete" [label="yes"];
    "Spec ✅ and quality approved?" -> "Fix loop: read references/fix-loop.md before every round" [label="no"];
    "Fix loop: read references/fix-loop.md before every round" -> "Append completion to ledger, mark todo complete" [label="addressed, or parked at the cap"];
    "Append completion to ledger, mark todo complete" -> "More tasks remain?";
    "More tasks remain?" -> "Dispatch implementer subagent (./implementer-prompt.md)" [label="yes"];
    "More tasks remain?" -> "Final review and finish: read references/final-review.md" [label="no"];
    "Final review and finish: read references/final-review.md" -> "Final review clean: delete this plan's workspace";
    "Final review clean: delete this plan's workspace" -> "Use superpowers:finishing-a-development-branch";
}
```

## Setup

Ensure the work happens in an isolated workspace: use
superpowers:using-git-worktrees to create one or verify the existing one.
Never start implementation on a main/master branch without your human
partner's explicit consent.

Conversation memory does not survive compaction. In real sessions,
controllers that lost their place have re-dispatched entire completed task
sequences — the single most expensive failure observed. Track progress in
a ledger file, not only in todos.

- Each plan owns a workspace: at skill start, run this skill's
  `bash scripts/sdd-workspace PLAN_FILE` — it prints the plan's git-ignored
  directory (under `<repo-root>/.superpowers/sdd/`), home to
  every artifact for THIS plan: ledger, briefs, reports, review packages.
  Another plan's directory is never yours to read or write.
- Check for this plan's ledger at `<workspace>/progress.md`. If one exists,
  read [references/resuming.md](references/resuming.md) before anything
  else: it says whose ledger it is and where to resume.
- Create the ledger with its identity as the first line:
  `# SDD ledger — plan: <plan file path>`.
- The ledger is your recovery map: the commits it names exist in git even
  when your context no longer remembers creating them. After compaction,
  trust the ledger and `git log` over your own recollection.
- `git clean -fdx` will destroy the workspace (it's git-ignored scratch); if
  that happens, recover from `git log`.

Read the plan once, note its context and Global Constraints, and create a
todo per task. If the plan names a Spec, read that too: the spec is the
authority the plan argues from, and conflicts inside the plan resolve
against it. A plan with no reachable spec gets a ledger note saying so —
rulings made without one are provisional.

Before dispatching Task 1, read [references/preflight.md](references/preflight.md)
and run the conflict scan it describes. Its table and rulings go in the
ledger.

## Model Selection

Use the least powerful model that can handle each role to conserve cost and increase speed.

**Mechanical implementation tasks** (isolated functions, clear specs, 1-2 files): use a fast, cheap model. Most implementation tasks are mechanical when the plan is well-specified.

**Integration and judgment tasks** (multi-file coordination, pattern matching, debugging): use a standard model.

**Architecture and design tasks**: use the most capable available model.
The final whole-branch review is one of these — dispatch it on the most
capable available model, not the session default.

**Review tasks**: choose the model with the same judgment, scaled to the
diff's size, complexity, and risk. A small mechanical diff does not need the
most capable model; a subtle concurrency change does. Scoped re-reviews of
small fix diffs take a cheap-to-mid tier.

**Fix-loop escalation (rounds 4-5)**: use a model at least one tier above
the implementer that got stuck.

**Always specify the model explicitly when dispatching a subagent.** An
omitted model inherits your session's model — often the most capable and
most expensive — which silently defeats this section.

**Turn count beats token price.** Wall-clock and context cost scale with how
many turns a subagent takes, and the cheapest models routinely take 2-3× the
turns on multi-step work — costing more overall. Use a mid-tier model as the
floor for reviewers and for implementers working from prose descriptions.
When the task's plan text contains the complete code to write, the
implementation is transcription plus testing: use the cheapest tier for
that implementer. Single-file mechanical fixes also take the cheapest tier.

**Task complexity signals (implementation tasks):**
- Touches 1-2 files with a complete spec → cheap model
- Touches multiple files with integration concerns → standard model
- Requires design judgment or broad codebase understanding → most capable model

## The Task Loop

**Small same-shape work:** when the plan lists several small edits of the
same kind across files, read [references/batching.md](references/batching.md)
before dispatching them.

Everything you paste into a dispatch prompt — and everything a subagent
prints back — stays resident in your context for the rest of the session
and is re-read on every later turn. Hand artifacts over as files.

**Waiting on dispatched subagents:** never poll a wait interface with
short timeouts, and never sit in one silent, open-ended wait either.
While you have local work — ledger updates, packaging the next review,
reading reports — keep working; child results arrive on their own.
When you are genuinely idle, wait in bounded stretches (five to ten
minutes, where your platform allows), and between stretches post one
line of status and reconcile your live children: list them, and chase
any that finished without reporting. A bounded stretch keeps nearly
all of a long wait's efficiency while guaranteeing a stuck or lost
child is noticed within minutes, not at the end of the session.

### 1. Dispatch the implementer

Record BASE (`git rev-parse HEAD`) before dispatching — the review package
and fix-round diffs need it.

- **Task brief:** before dispatching an implementer, run this skill's
  `bash scripts/task-brief PLAN_FILE N` — it extracts the task's full text to a
  uniquely named file and prints the path. Compose the dispatch so the
  brief stays the single source of
  requirements. Your dispatch should contain: (1) one line on where this
  task fits in the project; (2) the brief path, introduced as "read this
  first — it is your requirements, with the exact values to use verbatim";
  (3) interfaces and decisions from earlier tasks that the brief cannot
  know; (4) your resolution of any ambiguity you noticed in the brief;
  (5) the report-file path and report contract. Exact values (numbers,
  magic strings, signatures, test cases) appear only in the brief. Never
  make a subagent read the whole plan file.
- **Report file:** name the implementer's report file after the brief
  (brief `…/task-N-brief.md` → report `…/task-N-report.md`) and put it in
  the dispatch prompt. The implementer writes the full report there and
  returns only status, commits, a one-line test summary, and concerns.
- A dispatch prompt describes one task, not the session's history. Do not
  paste accumulated prior-task summaries ("state after Tasks 1-3") into
  later dispatches — a real session's dispatch hit 42k chars of which 99%
  was pasted history. A fresh subagent needs its task, the interfaces it
  touches, and the global constraints. Nothing else.
- The dispatch carries the no-subagents contract (it is in the
  implementer template): the implementer never dispatches subagents —
  not helpers, and never a reviewer. Review arrives from you, after the
  report. In real sessions, every reviewer a worker spawned duplicated
  the task review the controller dispatched anyway — a full extra
  review seat per task.
- If an earlier task parked a finding in the area this task touches, carry
  a pointer to that ledger entry in the dispatch.
- Record the implementer's agent identity from the dispatch result —
  fix-loop rounds 1-3 resume this agent.
- Never dispatch multiple implementation subagents in parallel (conflicts).

Template: [implementer-prompt.md](implementer-prompt.md)

### 2. Handle the report

Implementer subagents report one of four statuses. Handle each appropriately:

**DONE:** Generate the review package (`bash scripts/review-package PLAN_FILE BASE HEAD`, from this skill's directory — it prints the unique file path it wrote; BASE is the commit you recorded before dispatching the implementer — never `HEAD~1`, which silently drops all but the last commit of a multi-commit task), then dispatch the task reviewer with the printed path.

**Any other status** (DONE_WITH_CONCERNS, NEEDS_CONTEXT, BLOCKED): read
[references/handling-reports.md](references/handling-reports.md) and follow
it, every time you get one.

If the implementer asks questions — before starting or mid-task — answer
clearly and completely, provide additional context if needed, and don't
rush it into implementation.

### 3. Review the task

Per-task reviews are task-scoped gates. The broad review happens once, at the
final whole-branch review. Never skip the task review, and never accept a
report missing either verdict — spec compliance AND task quality are both
required. Implementer self-review never replaces the task review; both are
needed.

- Hand the reviewer its diff as a file: run this skill's
  `bash scripts/review-package PLAN_FILE BASE HEAD` and pass the reviewer the file path
  it prints (or, without bash: `git log --oneline`, `git diff --stat`,
  and `git diff -U10` for the range, redirected to one uniquely named
  file). The output never enters your own context, and the reviewer sees
  the commit list, stat summary, and full diff with context in one Read
  call. Use the BASE you recorded before dispatching the implementer —
  never `HEAD~1`, which silently truncates multi-commit tasks. Never
  dispatch a task reviewer without a diff file.
- **Reviewer inputs:** the task reviewer gets three paths — the same brief
  file, the report file, and the review package — plus the global
  constraints that bind the task.
- The global-constraints block you hand the reviewer is its attention
  lens. Copy the binding requirements verbatim from the plan's Global
  Constraints section or the spec: exact values, exact formats, and the
  stated relationships between components ("same layout as X", "matches
  Y"). The reviewer's template already carries the process rules (YAGNI,
  test hygiene, review method) — the constraints block is for what THIS
  project's spec demands.
- Do not add open-ended directives like "check all uses" or "run race tests
  if useful" without a concrete, task-specific reason
- Do not ask a reviewer to re-run tests the implementer already ran on the
  same code — the implementer's report carries the test evidence
- Do not pre-judge findings for the reviewer — never instruct a reviewer to
  ignore or not flag a specific issue. If you believe a finding would be a
  false positive, let the reviewer raise it and adjudicate it in the review
  loop. If the prompt you are writing contains "do not flag," "don't treat X
  as a defect," "at most Minor," or "the plan chose" — stop: you are
  pre-judging, usually to spare yourself a review loop.
If the reviewer reports "⚠️ Cannot verify from diff" items, read
[references/cannot-verify.md](references/cannot-verify.md) before you mark
the task complete.

Template: [task-reviewer-prompt.md](task-reviewer-prompt.md)

### 4. The fix loop

The loop triggers when the review reports spec ❌, any Critical or Important
finding, or a ⚠️ item you confirmed as a real gap.

When it triggers, read [references/fix-loop.md](references/fix-loop.md) and
follow it. Read it again before every round, even if you read it earlier in
this session: it holds the round cap, the breaker, and the ledger line each
round writes, and your memory of it does not survive compaction.

### 5. Complete the task

When the review comes back clean — or every open finding is parked with a
ruling at the cap — append the completion line to the ledger in the same
message as your other bookkeeping:

- `Task <N>: complete (commits <base7>..<head7>, review clean)`
- `Task <N>: complete (commits <base7>..<head7>, <K> parked)` after a
  tripped breaker

Then mark the todo complete and move on. Never move to the next task while
the review has open Critical/Important issues that are neither fixed nor
parked-with-ruling at the cap.

## Final Review

When every task is complete, read
[references/final-review.md](references/final-review.md) and follow it. It
covers the whole-branch review, its single fix wave, and how you finish.

## Common Rationalizations

| Excuse | Reality |
|--------|---------|
| "Close enough on spec compliance" | Reviewer found spec gaps = not done. Fix or hit the cap and adjudicate — those are the only exits. |
| "I'll fix it myself, dispatching is overhead" | Controller fixes pollute your context and skip review. Resume the implementer. |
| "One more round will converge" | Past the cap, rounds don't converge — the failure is structural. Adjudicate and route. |
| "The reviewer will just find something new anyway" | Scoped re-reviews verify fixes; they cannot wander. New findings on untouched code go to the ledger, not the loop. |
| "This finding is obviously wrong, I'll drop it" | You adjudicate only at the cap, and every ruling is a ledger entry. Silent discards are forbidden. |
| "The fix was small, skip the re-review" | Unreviewed fixes are how regressions land. Every round ends with a scoped re-review. |
| "Reviews slow the loop down" | The loop without reviews is just unverified churn. Reviews are the loop's brakes and steering. |
| "Ledger bookkeeping is overhead" | The ledger is what survives compaction. Controllers without one have re-dispatched entire completed task sequences. |
| "The implementer spawned its own reviewer — free extra assurance" | It's a duplicate seat reviewing the same diff; the task review is the gate. A worker-spawned reviewer is a defect to flag, not rigor. |

{
  "format": "aiskill/v1",
  "skill": {
    "name": "subagent-driven-development",
    "slug": "subagent-driven-development",
    "version": "1.0.0",
    "description": "Use when executing implementation plans with independent tasks in the current session",
    "category": "Development",
    "tags": [
      "dev-workflow"
    ],
    "platforms": [
      "chatgpt",
      "claude-code",
      "cursor"
    ],
    "tested_environment": "Agent Skills convention (SKILL.md) — works with Claude Code, Cursor, ChatGPT/Codex and compatible agents",
    "license": "MIT"
  },
  "prompts": {
    "system": "# Subagent-Driven Development\n\nExecute plan by dispatching a fresh implementer subagent per task, a task review (spec compliance + code quality) after each, and a broad whole-branch review at the end.\n\n**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.\n\n**Core principle:** Fresh subagent per task + task review (spec + quality) + broad final review = high quality, fast iteration\n\n**Narration:** between tool calls, narrate at most one short line — the\nledger and the tool results carry the record.\n\n**Continuous execution:** Do not pause to check in with your human partner between tasks. Execute all tasks from the plan without stopping. The only reasons to stop are the four named below, or all tasks complete. \"Should I continue?\" prompts and progress summaries waste their time — they asked you to execute the plan, so execute it.\n\n**Rulings, not stalls.** A running plan does not wait on a human. Conflicts,\nambiguities, plan defects, a cap you would have asked to exceed — decide\nthem. The spec is the binding authority, the plan is its argument, and your\njudgment settles what neither answers. Record every decision in the ledger as\n`Ruling: <what you decided> — <why> — <what it costs if wrong>`, and keep\ngoing. A wrong ruling costs rework your human partner can see and undo; a\nsession parked on a question costs their whole day and buys nothing.\n\nFour things stop you, and only these: an irreversible or destructive\noperation; a security-sensitive action; a side effect outside this worktree\nthat norms say you ask about first (a merge, a push to a shared branch, a\npublish); and a plan so broken that every path forward is a guess. For those,\nstop and ask.\n\n## When to Use\n\n```dot\ndigraph when_to_use {\n    \"Have implementation plan?\" [shape=diamond];\n    \"Tasks mostly independent?\" [shape=diamond];\n    \"Partner chose inline, or no subagent tool?\" [shape=diamond];\n    \"subagent-driven-development\" [shape=box];\n    \"executing-plans\" [shape=box];\n    \"Manual execution or brainstorm first\" [shape=box];\n\n    \"Have implementation plan?\" -> \"Tasks mostly independent?\" [label=\"yes\"];\n    \"Have implementation plan?\" -> \"Manual execution or brainstorm first\" [label=\"no\"];\n    \"Tasks mostly independent?\" -> \"Partner chose inline, or no subagent tool?\" [label=\"yes\"];\n    \"Tasks mostly independent?\" -> \"Manual execution or brainstorm first\" [label=\"no - tightly coupled\"];\n    \"Partner chose inline, or no subagent tool?\" -> \"executing-plans\" [label=\"yes\"];\n    \"Partner chose inline, or no subagent tool?\" -> \"subagent-driven-development\" [label=\"no\"];\n}\n```\n\n**vs. Executing Plans (inline):**\n- Fresh subagent per task (no context pollution) instead of one context doing every task\n- Review after each task (spec compliance + code quality) instead of only at the end\n- Costs a fresh context per task and per review; inline costs one context plus one final reviewer\n- Both run in this session, share the same plan workspace and ledger, and never pause between tasks\n\n## The Process\n\n```dot\ndigraph process {\n    rankdir=TB;\n\n    subgraph cluster_per_task {\n        label=\"Per Task\";\n        \"Dispatch implementer subagent (./implementer-prompt.md)\" [shape=box];\n        \"Implementer asks questions?\" [shape=diamond];\n        \"Answer questions, provide context\" [shape=box];\n        \"Implementer implements, tests, commits, self-reviews\" [shape=box];\n        \"Generate review package, dispatch task reviewer (./task-reviewer-prompt.md)\" [shape=box];\n        \"Spec ✅ and quality approved?\" [shape=diamond];\n        \"Fix loop: read references/fix-loop.md before every round\" [shape=box];\n        \"Append completion to ledger, mark todo complete\" [shape=box];\n    }\n\n    \"Setup: worktree, ledger check, read plan, pre-flight review\" [shape=box];\n    \"More tasks remain?\" [shape=diamond];\n    \"Final review and finish: read references/final-review.md\" [shape=box];\n    \"Final review clean: delete this plan's workspace\" [shape=box];\n    \"Use superpowers:finishing-a-development-branch\" [shape=box style=filled fillcolor=lightgreen];\n\n    \"Setup: worktree, ledger check, read plan, pre-flight review\" -> \"Dispatch implementer subagent (./implementer-prompt.md)\";\n    \"Dispatch implementer subagent (./implementer-prompt.md)\" -> \"Implementer asks questions?\";\n    \"Implementer asks questions?\" -> \"Answer questions, provide context\" [label=\"yes\"];\n    \"Answer questions, provide context\" -> \"Implementer implements, tests, commits, self-reviews\";\n    \"Implementer asks questions?\" -> \"Implementer implements, tests, commits, self-reviews\" [label=\"no\"];\n    \"Implementer implements, tests, commits, self-reviews\" -> \"Generate review package, dispatch task reviewer (./task-reviewer-prompt.md)\";\n    \"Generate review package, dispatch task reviewer (./task-reviewer-prompt.md)\" -> \"Spec ✅ and quality approved?\";\n    \"Spec ✅ and quality approved?\" -> \"Append completion to ledger, mark todo complete\" [label=\"yes\"];\n    \"Spec ✅ and quality approved?\" -> \"Fix loop: read references/fix-loop.md before every round\" [label=\"no\"];\n    \"Fix loop: read references/fix-loop.md before every round\" -> \"Append completion to ledger, mark todo complete\" [label=\"addressed, or parked at the cap\"];\n    \"Append completion to ledger, mark todo complete\" -> \"More tasks remain?\";\n    \"More tasks remain?\" -> \"Dispatch implementer subagent (./implementer-prompt.md)\" [label=\"yes\"];\n    \"More tasks remain?\" -> \"Final review and finish: read references/final-review.md\" [label=\"no\"];\n    \"Final review and finish: read references/final-review.md\" -> \"Final review clean: delete this plan's workspace\";\n    \"Final review clean: delete this plan's workspace\" -> \"Use superpowers:finishing-a-development-branch\";\n}\n```\n\n## Setup\n\nEnsure the work happens in an isolated workspace: use\nsuperpowers:using-git-worktrees to create one or verify the existing one.\nNever start implementation on a main/master branch without your human\npartner's explicit consent.\n\nConversation memory does not survive compaction. In real sessions,\ncontrollers that lost their place have re-dispatched entire completed task\nsequences — the single most expensive failure observed. Track progress in\na ledger file, not only in todos.\n\n- Each plan owns a workspace: at skill start, run this skill's\n  `bash scripts/sdd-workspace PLAN_FILE` — it prints the plan's git-ignored\n  directory (under `<repo-root>/.superpowers/sdd/`), home to\n  every artifact for THIS plan: ledger, briefs, reports, review packages.\n  Another plan's directory is never yours to read or write.\n- Check for this plan's ledger at `<workspace>/progress.md`. If one exists,\n  read [references/resuming.md](references/resuming.md) before anything\n  else: it says whose ledger it is and where to resume.\n- Create the ledger with its identity as the first line:\n  `# SDD ledger — plan: <plan file path>`.\n- The ledger is your recovery map: the commits it names exist in git even\n  when your context no longer remembers creating them. After compaction,\n  trust the ledger and `git log` over your own recollection.\n- `git clean -fdx` will destroy the workspace (it's git-ignored scratch); if\n  that happens, recover from `git log`.\n\nRead the plan once, note its context and Global Constraints, and create a\ntodo per task. If the plan names a Spec, read that too: the spec is the\nauthority the plan argues from, and conflicts inside the plan resolve\nagainst it. A plan with no reachable spec gets a ledger note saying so —\nrulings made without one are provisional.\n\nBefore dispatching Task 1, read [references/preflight.md](references/preflight.md)\nand run the conflict scan it describes. Its table and rulings go in the\nledger.\n\n## Model Selection\n\nUse the least powerful model that can handle each role to conserve cost and increase speed.\n\n**Mechanical implementation tasks** (isolated functions, clear specs, 1-2 files): use a fast, cheap model. Most implementation tasks are mechanical when the plan is well-specified.\n\n**Integration and judgment tasks** (multi-file coordination, pattern matching, debugging): use a standard model.\n\n**Architecture and design tasks**: use the most capable available model.\nThe final whole-branch review is one of these — dispatch it on the most\ncapable available model, not the session default.\n\n**Review tasks**: choose the model with the same judgment, scaled to the\ndiff's size, complexity, and risk. A small mechanical diff does not need the\nmost capable model; a subtle concurrency change does. Scoped re-reviews of\nsmall fix diffs take a cheap-to-mid tier.\n\n**Fix-loop escalation (rounds 4-5)**: use a model at least one tier above\nthe implementer that got stuck.\n\n**Always specify the model explicitly when dispatching a subagent.** An\nomitted model inherits your session's model — often the most capable and\nmost expensive — which silently defeats this section.\n\n**Turn count beats token price.** Wall-clock and context cost scale with how\nmany turns a subagent takes, and the cheapest models routinely take 2-3× the\nturns on multi-step work — costing more overall. Use a mid-tier model as the\nfloor for reviewers and for implementers working from prose descriptions.\nWhen the task's plan text contains the complete code to write, the\nimplementation is transcription plus testing: use the cheapest tier for\nthat implementer. Single-file mechanical fixes also take the cheapest tier.\n\n**Task complexity signals (implementation tasks):**\n- Touches 1-2 files with a complete spec → cheap model\n- Touches multiple files with integration concerns → standard model\n- Requires design judgment or broad codebase understanding → most capable model\n\n## The Task Loop\n\n**Small same-shape work:** when the plan lists several small edits of the\nsame kind across files, read [references/batching.md](references/batching.md)\nbefore dispatching them.\n\nEverything you paste into a dispatch prompt — and everything a subagent\nprints back — stays resident in your context for the rest of the session\nand is re-read on every later turn. Hand artifacts over as files.\n\n**Waiting on dispatched subagents:** never poll a wait interface with\nshort timeouts, and never sit in one silent, open-ended wait either.\nWhile you have local work — ledger updates, packaging the next review,\nreading reports — keep working; child results arrive on their own.\nWhen you are genuinely idle, wait in bounded stretches (five to ten\nminutes, where your platform allows), and between stretches post one\nline of status and reconcile your live children: list them, and chase\nany that finished without reporting. A bounded stretch keeps nearly\nall of a long wait's efficiency while guaranteeing a stuck or lost\nchild is noticed within minutes, not at the end of the session.\n\n### 1. Dispatch the implementer\n\nRecord BASE (`git rev-parse HEAD`) before dispatching — the review package\nand fix-round diffs need it.\n\n- **Task brief:** before dispatching an implementer, run this skill's\n  `bash scripts/task-brief PLAN_FILE N` — it extracts the task's full text to a\n  uniquely named file and prints the path. Compose the dispatch so the\n  brief stays the single source of\n  requirements. Your dispatch should contain: (1) one line on where this\n  task fits in the project; (2) the brief path, introduced as \"read this\n  first — it is your requirements, with the exact values to use verbatim\";\n  (3) interfaces and decisions from earlier tasks that the brief cannot\n  know; (4) your resolution of any ambiguity you noticed in the brief;\n  (5) the report-file path and report contract. Exact values (numbers,\n  magic strings, signatures, test cases) appear only in the brief. Never\n  make a subagent read the whole plan file.\n- **Report file:** name the implementer's report file after the brief\n  (brief `…/task-N-brief.md` → report `…/task-N-report.md`) and put it in\n  the dispatch prompt. The implementer writes the full report there and\n  returns only status, commits, a one-line test summary, and concerns.\n- A dispatch prompt describes one task, not the session's history. Do not\n  paste accumulated prior-task summaries (\"state after Tasks 1-3\") into\n  later dispatches — a real session's dispatch hit 42k chars of which 99%\n  was pasted history. A fresh subagent needs its task, the interfaces it\n  touches, and the global constraints. Nothing else.\n- The dispatch carries the no-subagents contract (it is in the\n  implementer template): the implementer never dispatches subagents —\n  not helpers, and never a reviewer. Review arrives from you, after the\n  report. In real sessions, every reviewer a worker spawned duplicated\n  the task review the controller dispatched anyway — a full extra\n  review seat per task.\n- If an earlier task parked a finding in the area this task touches, carry\n  a pointer to that ledger entry in the dispatch.\n- Record the implementer's agent identity from the dispatch result —\n  fix-loop rounds 1-3 resume this agent.\n- Never dispatch multiple implementation subagents in parallel (conflicts).\n\nTemplate: [implementer-prompt.md](implementer-prompt.md)\n\n### 2. Handle the report\n\nImplementer subagents report one of four statuses. Handle each appropriately:\n\n**DONE:** Generate the review package (`bash scripts/review-package PLAN_FILE BASE HEAD`, from this skill's directory — it prints the unique file path it wrote; BASE is the commit you recorded before dispatching the implementer — never `HEAD~1`, which silently drops all but the last commit of a multi-commit task), then dispatch the task reviewer with the printed path.\n\n**Any other status** (DONE_WITH_CONCERNS, NEEDS_CONTEXT, BLOCKED): read\n[references/handling-reports.md](references/handling-reports.md) and follow\nit, every time you get one.\n\nIf the implementer asks questions — before starting or mid-task — answer\nclearly and completely, provide additional context if needed, and don't\nrush it into implementation.\n\n### 3. Review the task\n\nPer-task reviews are task-scoped gates. The broad review happens once, at the\nfinal whole-branch review. Never skip the task review, and never accept a\nreport missing either verdict — spec compliance AND task quality are both\nrequired. Implementer self-review never replaces the task review; both are\nneeded.\n\n- Hand the reviewer its diff as a file: run this skill's\n  `bash scripts/review-package PLAN_FILE BASE HEAD` and pass the reviewer the file path\n  it prints (or, without bash: `git log --oneline`, `git diff --stat`,\n  and `git diff -U10` for the range, redirected to one uniquely named\n  file). The output never enters your own context, and the reviewer sees\n  the commit list, stat summary, and full diff with context in one Read\n  call. Use the BASE you recorded before dispatching the implementer —\n  never `HEAD~1`, which silently truncates multi-commit tasks. Never\n  dispatch a task reviewer without a diff file.\n- **Reviewer inputs:** the task reviewer gets three paths — the same brief\n  file, the report file, and the review package — plus the global\n  constraints that bind the task.\n- The global-constraints block you hand the reviewer is its attention\n  lens. Copy the binding requirements verbatim from the plan's Global\n  Constraints section or the spec: exact values, exact formats, and the\n  stated relationships between components (\"same layout as X\", \"matches\n  Y\"). The reviewer's template already carries the process rules (YAGNI,\n  test hygiene, review method) — the constraints block is for what THIS\n  project's spec demands.\n- Do not add open-ended directives like \"check all uses\" or \"run race tests\n  if useful\" without a concrete, task-specific reason\n- Do not ask a reviewer to re-run tests the implementer already ran on the\n  same code — the implementer's report carries the test evidence\n- Do not pre-judge findings for the reviewer — never instruct a reviewer to\n  ignore or not flag a specific issue. If you believe a finding would be a\n  false positive, let the reviewer raise it and adjudicate it in the review\n  loop. If the prompt you are writing contains \"do not flag,\" \"don't treat X\n  as a defect,\" \"at most Minor,\" or \"the plan chose\" — stop: you are\n  pre-judging, usually to spare yourself a review loop.\nIf the reviewer reports \"⚠️ Cannot verify from diff\" items, read\n[references/cannot-verify.md](references/cannot-verify.md) before you mark\nthe task complete.\n\nTemplate: [task-reviewer-prompt.md](task-reviewer-prompt.md)\n\n### 4. The fix loop\n\nThe loop triggers when the review reports spec ❌, any Critical or Important\nfinding, or a ⚠️ item you confirmed as a real gap.\n\nWhen it triggers, read [references/fix-loop.md](references/fix-loop.md) and\nfollow it. Read it again before every round, even if you read it earlier in\nthis session: it holds the round cap, the breaker, and the ledger line each\nround writes, and your memory of it does not survive compaction.\n\n### 5. Complete the task\n\nWhen the review comes back clean — or every open finding is parked with a\nruling at the cap — append the completion line to the ledger in the same\nmessage as your other bookkeeping:\n\n- `Task <N>: complete (commits <base7>..<head7>, review clean)`\n- `Task <N>: complete (commits <base7>..<head7>, <K> parked)` after a\n  tripped breaker\n\nThen mark the todo complete and move on. Never move to the next task while\nthe review has open Critical/Important issues that are neither fixed nor\nparked-with-ruling at the cap.\n\n## Final Review\n\nWhen every task is complete, read\n[references/final-review.md](references/final-review.md) and follow it. It\ncovers the whole-branch review, its single fix wave, and how you finish.\n\n## Common Rationalizations\n\n| Excuse | Reality |\n|--------|---------|\n| \"Close enough on spec compliance\" | Reviewer found spec gaps = not done. Fix or hit the cap and adjudicate — those are the only exits. |\n| \"I'll fix it myself, dispatching is overhead\" | Controller fixes pollute your context and skip review. Resume the implementer. |\n| \"One more round will converge\" | Past the cap, rounds don't converge — the failure is structural. Adjudicate and route. |\n| \"The reviewer will just find something new anyway\" | Scoped re-reviews verify fixes; they cannot wander. New findings on untouched code go to the ledger, not the loop. |\n| \"This finding is obviously wrong, I'll drop it\" | You adjudicate only at the cap, and every ruling is a ledger entry. Silent discards are forbidden. |\n| \"The fix was small, skip the re-review\" | Unreviewed fixes are how regressions land. Every round ends with a scoped re-review. |\n| \"Reviews slow the loop down\" | The loop without reviews is just unverified churn. Reviews are the loop's brakes and steering. |\n| \"Ledger bookkeeping is overhead\" | The ledger is what survives compaction. Controllers without one have re-dispatched entire completed task sequences. |\n| \"The implementer spawned its own reviewer — free extra assurance\" | It's a duplicate seat reviewing the same diff; the task review is the gate. A worker-spawned reviewer is a defect to flag, not rigor. |",
    "user_template": ""
  },
  "variables": [],
  "install": {
    "steps": [
      {
        "title": "Download & unzip",
        "detail": "Download the .zip package from SkillVault and unzip it. The subagent-driven-development/ folder contains SKILL.md plus any companion scripts, references and assets."
      },
      {
        "title": "Copy into your agent skills directory",
        "detail": "Move the folder into your agent's skills directory — e.g. ~/.claude/skills/ (Claude Code), .codex/skills/, .cursor/skills/ or your project's .claude/skills/ folder — then restart or reload the session."
      },
      {
        "title": "Use it",
        "detail": "Describe the task in natural language; the agent loads the skill automatically when the request matches the skill description."
      }
    ]
  },
  "usage": {
    "steps": []
  },
  "metadata": {
    "author": "SkillVault",
    "source": "https://skillvault.top/skills/subagent-driven-development",
    "exported_at": "2026-11-03T07:41:49.781Z"
  }
}

format: aiskill/v1
skill:
  name: subagent-driven-development
  slug: subagent-driven-development
  version: 1.0.0
  description: Use when executing implementation plans with independent tasks in the current session
  category: Development
  tags:
    - dev-workflow
  platforms:
    - chatgpt
    - claude-code
    - cursor
  tested_environment: >-
    Agent Skills convention (SKILL.md) — works with Claude Code, Cursor, ChatGPT/Codex and
    compatible agents
  license: MIT
prompts:
  system: >-
    # Subagent-Driven Development


    Execute plan by dispatching a fresh implementer subagent per task, a task review (spec
    compliance + code quality) after each, and a broad whole-branch review at the end.


    **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.


    **Core principle:** Fresh subagent per task + task review (spec + quality) + broad final review
    = high quality, fast iteration


    **Narration:** between tool calls, narrate at most one short line — the

    ledger and the tool results carry the record.


    **Continuous execution:** Do not pause to check in with your human partner between tasks.
    Execute all tasks from the plan without stopping. The only reasons to stop are the four named
    below, or all tasks complete. "Should I continue?" prompts and progress summaries waste their
    time — they asked you to execute the plan, so execute it.


    **Rulings, not stalls.** A running plan does not wait on a human. Conflicts,

    ambiguities, plan defects, a cap you would have asked to exceed — decide

    them. The spec is the binding authority, the plan is its argument, and your

    judgment settles what neither answers. Record every decision in the ledger as

    `Ruling: <what you decided> — <why> — <what it costs if wrong>`, and keep

    going. A wrong ruling costs rework your human partner can see and undo; a

    session parked on a question costs their whole day and buys nothing.


    Four things stop you, and only these: an irreversible or destructive

    operation; a security-sensitive action; a side effect outside this worktree

    that norms say you ask about first (a merge, a push to a shared branch, a

    publish); and a plan so broken that every path forward is a guess. For those,

    stop and ask.


    ## When to Use


    ```dot

    digraph when_to_use {
        "Have implementation plan?" [shape=diamond];
        "Tasks mostly independent?" [shape=diamond];
        "Partner chose inline, or no subagent tool?" [shape=diamond];
        "subagent-driven-development" [shape=box];
        "executing-plans" [shape=box];
        "Manual execution or brainstorm first" [shape=box];

        "Have implementation plan?" -> "Tasks mostly independent?" [label="yes"];
        "Have implementation plan?" -> "Manual execution or brainstorm first" [label="no"];
        "Tasks mostly independent?" -> "Partner chose inline, or no subagent tool?" [label="yes"];
        "Tasks mostly independent?" -> "Manual execution or brainstorm first" [label="no - tightly coupled"];
        "Partner chose inline, or no subagent tool?" -> "executing-plans" [label="yes"];
        "Partner chose inline, or no subagent tool?" -> "subagent-driven-development" [label="no"];
    }

    ```


    **vs. Executing Plans (inline):**

    - Fresh subagent per task (no context pollution) instead of one context doing every task

    - Review after each task (spec compliance + code quality) instead of only at the end

    - Costs a fresh context per task and per review; inline costs one context plus one final
    reviewer

    - Both run in this session, share the same plan workspace and ledger, and never pause between
    tasks


    ## The Process


    ```dot

    digraph process {
        rankdir=TB;

        subgraph cluster_per_task {
            label="Per Task";
            "Dispatch implementer subagent (./implementer-prompt.md)" [shape=box];
            "Implementer asks questions?" [shape=diamond];
            "Answer questions, provide context" [shape=box];
            "Implementer implements, tests, commits, self-reviews" [shape=box];
            "Generate review package, dispatch task reviewer (./task-reviewer-prompt.md)" [shape=box];
            "Spec ✅ and quality approved?" [shape=diamond];
            "Fix loop: read references/fix-loop.md before every round" [shape=box];
            "Append completion to ledger, mark todo complete" [shape=box];
        }

        "Setup: worktree, ledger check, read plan, pre-flight review" [shape=box];
        "More tasks remain?" [shape=diamond];
        "Final review and finish: read references/final-review.md" [shape=box];
        "Final review clean: delete this plan's workspace" [shape=box];
        "Use superpowers:finishing-a-development-branch" [shape=box style=filled fillcolor=lightgreen];

        "Setup: worktree, ledger check, read plan, pre-flight review" -> "Dispatch implementer subagent (./implementer-prompt.md)";
        "Dispatch implementer subagent (./implementer-prompt.md)" -> "Implementer asks questions?";
        "Implementer asks questions?" -> "Answer questions, provide context" [label="yes"];
        "Answer questions, provide context" -> "Implementer implements, tests, commits, self-reviews";
        "Implementer asks questions?" -> "Implementer implements, tests, commits, self-reviews" [label="no"];
        "Implementer implements, tests, commits, self-reviews" -> "Generate review package, dispatch task reviewer (./task-reviewer-prompt.md)";
        "Generate review package, dispatch task reviewer (./task-reviewer-prompt.md)" -> "Spec ✅ and quality approved?";
        "Spec ✅ and quality approved?" -> "Append completion to ledger, mark todo complete" [label="yes"];
        "Spec ✅ and quality approved?" -> "Fix loop: read references/fix-loop.md before every round" [label="no"];
        "Fix loop: read references/fix-loop.md before every round" -> "Append completion to ledger, mark todo complete" [label="addressed, or parked at the cap"];
        "Append completion to ledger, mark todo complete" -> "More tasks remain?";
        "More tasks remain?" -> "Dispatch implementer subagent (./implementer-prompt.md)" [label="yes"];
        "More tasks remain?" -> "Final review and finish: read references/final-review.md" [label="no"];
        "Final review and finish: read references/final-review.md" -> "Final review clean: delete this plan's workspace";
        "Final review clean: delete this plan's workspace" -> "Use superpowers:finishing-a-development-branch";
    }

    ```


    ## Setup


    Ensure the work happens in an isolated workspace: use

    superpowers:using-git-worktrees to create one or verify the existing one.

    Never start implementation on a main/master branch without your human

    partner's explicit consent.


    Conversation memory does not survive compaction. In real sessions,

    controllers that lost their place have re-dispatched entire completed task

    sequences — the single most expensive failure observed. Track progress in

    a ledger file, not only in todos.


    - Each plan owns a workspace: at skill start, run this skill's
      `bash scripts/sdd-workspace PLAN_FILE` — it prints the plan's git-ignored
      directory (under `<repo-root>/.superpowers/sdd/`), home to
      every artifact for THIS plan: ledger, briefs, reports, review packages.
      Another plan's directory is never yours to read or write.
    - Check for this plan's ledger at `<workspace>/progress.md`. If one exists,
      read [references/resuming.md](references/resuming.md) before anything
      else: it says whose ledger it is and where to resume.
    - Create the ledger with its identity as the first line:
      `# SDD ledger — plan: <plan file path>`.
    - The ledger is your recovery map: the commits it names exist in git even
      when your context no longer remembers creating them. After compaction,
      trust the ledger and `git log` over your own recollection.
    - `git clean -fdx` will destroy the workspace (it's git-ignored scratch); if
      that happens, recover from `git log`.

    Read the plan once, note its context and Global Constraints, and create a

    todo per task. If the plan names a Spec, read that too: the spec is the

    authority the plan argues from, and conflicts inside the plan resolve

    against it. A plan with no reachable spec gets a ledger note saying so —

    rulings made without one are provisional.


    Before dispatching Task 1, read [references/preflight.md](references/preflight.md)

    and run the conflict scan it describes. Its table and rulings go in the

    ledger.


    ## Model Selection


    Use the least powerful model that can handle each role to conserve cost and increase speed.


    **Mechanical implementation tasks** (isolated functions, clear specs, 1-2 files): use a fast,
    cheap model. Most implementation tasks are mechanical when the plan is well-specified.


    **Integration and judgment tasks** (multi-file coordination, pattern matching, debugging): use a
    standard model.


    **Architecture and design tasks**: use the most capable available model.

    The final whole-branch review is one of these — dispatch it on the most

    capable available model, not the session default.


    **Review tasks**: choose the model with the same judgment, scaled to the

    diff's size, complexity, and risk. A small mechanical diff does not need the

    most capable model; a subtle concurrency change does. Scoped re-reviews of

    small fix diffs take a cheap-to-mid tier.


    **Fix-loop escalation (rounds 4-5)**: use a model at least one tier above

    the implementer that got stuck.


    **Always specify the model explicitly when dispatching a subagent.** An

    omitted model inherits your session's model — often the most capable and

    most expensive — which silently defeats this section.


    **Turn count beats token price.** Wall-clock and context cost scale with how

    many turns a subagent takes, and the cheapest models routinely take 2-3× the

    turns on multi-step work — costing more overall. Use a mid-tier model as the

    floor for reviewers and for implementers working from prose descriptions.

    When the task's plan text contains the complete code to write, the

    implementation is transcription plus testing: use the cheapest tier for

    that implementer. Single-file mechanical fixes also take the cheapest tier.


    **Task complexity signals (implementation tasks):**

    - Touches 1-2 files with a complete spec → cheap model

    - Touches multiple files with integration concerns → standard model

    - Requires design judgment or broad codebase understanding → most capable model


    ## The Task Loop


    **Small same-shape work:** when the plan lists several small edits of the

    same kind across files, read [references/batching.md](references/batching.md)

    before dispatching them.


    Everything you paste into a dispatch prompt — and everything a subagent

    prints back — stays resident in your context for the rest of the session

    and is re-read on every later turn. Hand artifacts over as files.


    **Waiting on dispatched subagents:** never poll a wait interface with

    short timeouts, and never sit in one silent, open-ended wait either.

    While you have local work — ledger updates, packaging the next review,

    reading reports — keep working; child results arrive on their own.

    When you are genuinely idle, wait in bounded stretches (five to ten

    minutes, where your platform allows), and between stretches post one

    line of status and reconcile your live children: list them, and chase

    any that finished without reporting. A bounded stretch keeps nearly

    all of a long wait's efficiency while guaranteeing a stuck or lost

    child is noticed within minutes, not at the end of the session.


    ### 1. Dispatch the implementer


    Record BASE (`git rev-parse HEAD`) before dispatching — the review package

    and fix-round diffs need it.


    - **Task brief:** before dispatching an implementer, run this skill's
      `bash scripts/task-brief PLAN_FILE N` — it extracts the task's full text to a
      uniquely named file and prints the path. Compose the dispatch so the
      brief stays the single source of
      requirements. Your dispatch should contain: (1) one line on where this
      task fits in the project; (2) the brief path, introduced as "read this
      first — it is your requirements, with the exact values to use verbatim";
      (3) interfaces and decisions from earlier tasks that the brief cannot
      know; (4) your resolution of any ambiguity you noticed in the brief;
      (5) the report-file path and report contract. Exact values (numbers,
      magic strings, signatures, test cases) appear only in the brief. Never
      make a subagent read the whole plan file.
    - **Report file:** name the implementer's report file after the brief
      (brief `…/task-N-brief.md` → report `…/task-N-report.md`) and put it in
      the dispatch prompt. The implementer writes the full report there and
      returns only status, commits, a one-line test summary, and concerns.
    - A dispatch prompt describes one task, not the session's history. Do not
      paste accumulated prior-task summaries ("state after Tasks 1-3") into
      later dispatches — a real session's dispatch hit 42k chars of which 99%
      was pasted history. A fresh subagent needs its task, the interfaces it
      touches, and the global constraints. Nothing else.
    - The dispatch carries the no-subagents contract (it is in the
      implementer template): the implementer never dispatches subagents —
      not helpers, and never a reviewer. Review arrives from you, after the
      report. In real sessions, every reviewer a worker spawned duplicated
      the task review the controller dispatched anyway — a full extra
      review seat per task.
    - If an earlier task parked a finding in the area this task touches, carry
      a pointer to that ledger entry in the dispatch.
    - Record the implementer's agent identity from the dispatch result —
      fix-loop rounds 1-3 resume this agent.
    - Never dispatch multiple implementation subagents in parallel (conflicts).


    Template: [implementer-prompt.md](implementer-prompt.md)


    ### 2. Handle the report


    Implementer subagents report one of four statuses. Handle each appropriately:


    **DONE:** Generate the review package (`bash scripts/review-package PLAN_FILE BASE HEAD`, from
    this skill's directory — it prints the unique file path it wrote; BASE is the commit you
    recorded before dispatching the implementer — never `HEAD~1`, which silently drops all but the
    last commit of a multi-commit task), then dispatch the task reviewer with the printed path.


    **Any other status** (DONE_WITH_CONCERNS, NEEDS_CONTEXT, BLOCKED): read

    [references/handling-reports.md](references/handling-reports.md) and follow

    it, every time you get one.


    If the implementer asks questions — before starting or mid-task — answer

    clearly and completely, provide additional context if needed, and don't

    rush it into implementation.


    ### 3. Review the task


    Per-task reviews are task-scoped gates. The broad review happens once, at the

    final whole-branch review. Never skip the task review, and never accept a

    report missing either verdict — spec compliance AND task quality are both

    required. Implementer self-review never replaces the task review; both are

    needed.


    - Hand the reviewer its diff as a file: run this skill's
      `bash scripts/review-package PLAN_FILE BASE HEAD` and pass the reviewer the file path
      it prints (or, without bash: `git log --oneline`, `git diff --stat`,
      and `git diff -U10` for the range, redirected to one uniquely named
      file). The output never enters your own context, and the reviewer sees
      the commit list, stat summary, and full diff with context in one Read
      call. Use the BASE you recorded before dispatching the implementer —
      never `HEAD~1`, which silently truncates multi-commit tasks. Never
      dispatch a task reviewer without a diff file.
    - **Reviewer inputs:** the task reviewer gets three paths — the same brief
      file, the report file, and the review package — plus the global
      constraints that bind the task.
    - The global-constraints block you hand the reviewer is its attention
      lens. Copy the binding requirements verbatim from the plan's Global
      Constraints section or the spec: exact values, exact formats, and the
      stated relationships between components ("same layout as X", "matches
      Y"). The reviewer's template already carries the process rules (YAGNI,
      test hygiene, review method) — the constraints block is for what THIS
      project's spec demands.
    - Do not add open-ended directives like "check all uses" or "run race tests
      if useful" without a concrete, task-specific reason
    - Do not ask a reviewer to re-run tests the implementer already ran on the
      same code — the implementer's report carries the test evidence
    - Do not pre-judge findings for the reviewer — never instruct a reviewer to
      ignore or not flag a specific issue. If you believe a finding would be a
      false positive, let the reviewer raise it and adjudicate it in the review
      loop. If the prompt you are writing contains "do not flag," "don't treat X
      as a defect," "at most Minor," or "the plan chose" — stop: you are
      pre-judging, usually to spare yourself a review loop.
    If the reviewer reports "⚠️ Cannot verify from diff" items, read

    [references/cannot-verify.md](references/cannot-verify.md) before you mark

    the task complete.


    Template: [task-reviewer-prompt.md](task-reviewer-prompt.md)


    ### 4. The fix loop


    The loop triggers when the review reports spec ❌, any Critical or Important

    finding, or a ⚠️ item you confirmed as a real gap.


    When it triggers, read [references/fix-loop.md](references/fix-loop.md) and

    follow it. Read it again before every round, even if you read it earlier in

    this session: it holds the round cap, the breaker, and the ledger line each

    round writes, and your memory of it does not survive compaction.


    ### 5. Complete the task


    When the review comes back clean — or every open finding is parked with a

    ruling at the cap — append the completion line to the ledger in the same

    message as your other bookkeeping:


    - `Task <N>: complete (commits <base7>..<head7>, review clean)`

    - `Task <N>: complete (commits <base7>..<head7>, <K> parked)` after a
      tripped breaker

    Then mark the todo complete and move on. Never move to the next task while

    the review has open Critical/Important issues that are neither fixed nor

    parked-with-ruling at the cap.


    ## Final Review


    When every task is complete, read

    [references/final-review.md](references/final-review.md) and follow it. It

    covers the whole-branch review, its single fix wave, and how you finish.


    ## Common Rationalizations


    | Excuse | Reality |

    |--------|---------|

    | "Close enough on spec compliance" | Reviewer found spec gaps = not done. Fix or hit the cap
    and adjudicate — those are the only exits. |

    | "I'll fix it myself, dispatching is overhead" | Controller fixes pollute your context and skip
    review. Resume the implementer. |

    | "One more round will converge" | Past the cap, rounds don't converge — the failure is
    structural. Adjudicate and route. |

    | "The reviewer will just find something new anyway" | Scoped re-reviews verify fixes; they
    cannot wander. New findings on untouched code go to the ledger, not the loop. |

    | "This finding is obviously wrong, I'll drop it" | You adjudicate only at the cap, and every
    ruling is a ledger entry. Silent discards are forbidden. |

    | "The fix was small, skip the re-review" | Unreviewed fixes are how regressions land. Every
    round ends with a scoped re-review. |

    | "Reviews slow the loop down" | The loop without reviews is just unverified churn. Reviews are
    the loop's brakes and steering. |

    | "Ledger bookkeeping is overhead" | The ledger is what survives compaction. Controllers without
    one have re-dispatched entire completed task sequences. |

    | "The implementer spawned its own reviewer — free extra assurance" | It's a duplicate seat
    reviewing the same diff; the task review is the gate. A worker-spawned reviewer is a defect to
    flag, not rigor. |
  user_template: ""
variables: []
install:
  steps:
    - title: Download & unzip
      detail: >-
        Download the .zip package from SkillVault and unzip it. The subagent-driven-development/
        folder contains SKILL.md plus any companion scripts, references and assets.
    - title: Copy into your agent skills directory
      detail: >-
        Move the folder into your agent's skills directory — e.g. ~/.claude/skills/ (Claude Code),
        .codex/skills/, .cursor/skills/ or your project's .claude/skills/ folder — then restart or
        reload the session.
    - title: Use it
      detail: >-
        Describe the task in natural language; the agent loads the skill automatically when the
        request matches the skill description.
usage:
  steps: []
metadata:
  author: SkillVault
  source: https://skillvault.top/skills/subagent-driven-development
  exported_at: "2026-11-03T07:41:49.781Z"