start

A command that starts an autonomous coding session for a stated goal. It can work through pull requests and merge them automatically.

In plain words
What is it for?
Use it to give an agent a development goal and control its sessions, pull requests, parallel work, models, and merge behavior.
Why use it?
It reduces the need to repeatedly supervise planning, implementation, fixes, and merging.

Command

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/developerz-ai/ai-task-master/start
Clone the repo
git clone --depth 1 https://github.com/developerz-ai/ai-task-master
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 3,288 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 $0.03288
Opus 5 $0.00000 $0.01644
Sonnet 5 $0.00000 $0.00658
Haiku 4.5 $0.00000 $0.00329

Measured yesterday against content hash 4682e3a3cb01, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

start 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 yesterday.

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.

packages/aitm/docs/commands/start.md · 200 lines

How it starts

The opening of the file, as written. The whole thing — 200 lines — stays where its author put it; the contents beside it link to each section on GitHub.

aitm start

Kick off an autonomous work session against a goal. With automerge on (default) the run drives every PR to merge by itself.

Signature

aitm start "<goal>"
  [--criteria "..."]
  [--max-prs N]            # default unbounded (0 = unbounded)
  [--max-sessions N]       # default unlimited
  [--max-fix-attempts N]   # default 3 — CI-fix passes per PR group before it blocks
  [--concurrency N]        # default 1 — PR groups worked in parallel
  [--subagents N]          # default 10 — subagents one lead may run at once
  [--no-automerge]         # default: automerge on
  [--admin]                # merge past base-branch protection (gh pr merge --admin)
  [--style <path>]         # default: detected CLAUDE.md or AGENTS.md
  [--model <id>]           # default: provider default
  [--branch <name>]        # default: aitm/<group-id>-<title-slug>

--max-fix-attempts and --concurrency override the persisted maxCiFixAttempts / concurrency config keys for this run; --no-automerge, --style, and --model likewise override their config counterparts. --admin is CLI-only — a per-run force-merge with no config-file key. The many run settings that have no flag at all (e.g. verifyCommand, selfReview, webSearch) are documented in config.md.

--max-fix-attempts

Caps how many CI-fix passes a PR group gets before it blocks for a human instead of looping on an unfixable red PR. Overrides the maxCiFixAttempts config key for this run (default 3). Each attempt is a coding-tier fix session plus a remote CI round-trip, so it is a direct cost/patience knob. See config.md.

--concurrency

How many PR groups may have a Worker running at the same time (default 1, sequential). Overrides the concurrency config key for this run. See config.md.

--admin

Merge with gh pr merge --admin to land a green PR past a base-branch protection rule that would otherwise block a solo-authored merge. It overrides the policy only — it never skips CI or merges failing checks (those still route to the CI-fix loop). Needed when running aitm against a repo whose main requires an approving review. See the repo CLAUDE.md §"Branch protection".

Read the full file on GitHub · 200 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. yesterday First seen · 200 lines · 0 tokens per session scan A 4682e3a3cb01

Subscribe to this mod's changes

start is a command published in the GitHub repository developerz-ai/ai-task-master (5 stars, last pushed 25d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,288 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 2026-08-31.