start-task

A workflow executor for completing an existing software task made up of smaller missions. Each mission is handled by a separate helper process using TDD, or test-driven development, and the results are checked and committed.

In plain words
What is it for?
Running remaining missions, handling vague requirements before execution, verifying tests after each mission, committing completed work, and reviewing the final changes.
Why use it?
It turns a multi-step task into a repeatable run with progress tracking, tests, and a separate Git commit for each completed mission. It starts immediately and pauses only when work or checks fail.

Command for Claude Code

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 commands/ayoubben18/ab-method/start-task
Clone the repo
git clone --depth 1 https://github.com/ayoubben18/ab-method

Made for: Claude Code.

Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 688 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 $0.00000 $0.00688
Opus 5 $0.00000 $0.00344
Sonnet 5 $0.00000 $0.00138
Haiku 4.5 $0.00000 $0.00069

Measured 2d ago against content hash ccd87e6b11cd, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

start-task 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 2d 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.

.claude/commands/start-task.md · 50 lines

How it starts

The opening of the file, as written. The whole thing — 50 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Start Task

Description

Run an existing task autonomously to completion — /resume-task with the /goal philosophy. Each remaining mission runs in a subagent that follows tdd and updates the progress tracker itself; the parent verifies tests and commits after every finished mission. Starts immediately on invocation — no "Proceed?" prompt — then no prompts until done (or something goes red).

Usage

/start-task [task-name]

Behavior

Loads and executes the start-task workflow from .ab-method/core/start-task.md

This workflow will:

  1. Identify the task and read its progress-tracker.md (single source of truth)
  2. Gate on vagueness — any mission too fuzzy to run without judgment calls gets grilled before the run, never mid-run
  3. Announce the plan (missions, parallel groups, commit-per-mission) and start immediately — no "Proceed?"; invoking the command is the consent
  4. Run every remaining mission in a subagent through the tdd discipline; the subagent checks off its mission and appends its technical summary to the tracker
  5. Verify the test suite after each mission, then commit (one commit per mission, one per [pp-x] group)
  6. Run review-implementation after the last mission (autonomous mode) — three critics (cleaner-architecture, slop-defender, reusability-inspector) on the task diff; safe fixes auto-applied (tests-green-gated, own commit), everything written to docs/tasks/<task>/review.md for you to read afk
  7. Stop loudly on red — never commits broken work, never starts the next mission on a broken state

/start-task vs /resume-task

  • /start-task — trust the roadmap, walk away, review commits instead of missions
  • /resume-task — stay in the loop, review each mission before moving on

Workflow Details

  • No confirmation — starts immediately, then autonomous — no "Proceed?" gate, no prompting between missions
  • Group-aware[pp-x] missions run concurrently in subagents; siblings skip the tracker (the parent merges their summaries) to avoid write conflicts
  • Green tests gate every commit — a red feedback loop takes priority over progress, exactly like /goal
  • Executor, not producer — missions are defined by /create-task / /extend-task; this workflow only runs them
  • A black box doesn't stop the run — a mission marked ⚠️ UQ-n builds the placeholder recorded in unresolved-questions.md and keeps going; the run never answers a parked question, and every open one it built on is listed in the announcement and the final report

Read the full file on GitHub · 50 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. 2d ago First seen · 50 lines · 0 tokens per session scan A ccd87e6b11cd

Subscribe to this mod's changes

start-task is a command published in the GitHub repository ayoubben18/ab-method (187 stars, last pushed 1mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 688 tokens. 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.