auto-iter

A command that runs one autonomous software iteration cycle. It selects a mode from the project backlog, performs the documented work, and handles safe or risky stopping points.

In plain words
What is it for?
Choosing between implementation, architecture review, brainstorming, and stopping; dispatching coding or review tasks; and applying the project’s pre-merge checks.
Why use it?
It coordinates the complete cycle without requiring the developer to run each phase manually.

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/proggarapsody/bitbottle/auto-iter
Clone the repo
git clone --depth 1 https://github.com/proggarapsody/bitbottle

Made for: Claude Code.

Per session 57 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,183 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 $0.00057 $0.02183
Opus 5 $0.00028 $0.01092
Sonnet 5 $0.00011 $0.00437
Haiku 4.5 $0.00006 $0.00218

Measured 2d ago against content hash 01e2220c7ef0, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

auto-iter 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/auto-iter.md · 147 lines

How it starts

The opening of the file, as written. The whole thing — 147 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/auto-iter — Claude Code adapter

Full iteration procedure: docs/workflows/iteration-cycle/README.md Autonomous-mode deltas: docs/workflows/iteration-cycle/autonomous.md Pre-merge gate: docs/workflows/pre-merge-check.md Quickref (tables + policies): docs/workflows/iteration-cycle/quickref.md Script catalog: docs/workflows/iteration-cycle/scripts.md

Model tier mapping

Phase tag (from quickref) Claude model
judgment-heavy Opus
code-generation Sonnet (subagent via Task tool)
mechanical Sonnet inline

Subagent dispatch

All heavy work dispatches via the Task tool:

Phase Dispatch Model Returns
§3 TDD implementation Task(isolation: "worktree") Sonnet subagent commit list + subagent_tokens
§6 design-judge Task(model: "sonnet") Sonnet subagent findings_count + subagent_tokens
Architecture audit Task(model: "opus") Opus subagent finding or audit_no_findings
Brainstorm Task(model: "opus") Opus subagent new BACKLOG rows

Orchestrator-inline (NOT dispatched): §0a lock, §0 preflight, §1 mode/scope pick, §1 bundle-check, §2 worktree creation/removal, §2 push PR + poll CI + merge, §4 log writes, §5 release lock.

If a step listed above as dispatched is run inline by mistake, record dispatch_violation: true in the metrics line so post-cycle analysis catches it.

In-cycle parallel block (§3.6)

After the §3.5 mechanical taste-check passes (autonomous.md §3.5), three activities run concurrently inside one cycle:

  1. Push branch + open PR (orchestrator-inline) — git push -u, then gh pr create with the PR body (see § "PRD-close enforcement" below for the Closes #<PRD> requirement).
  2. Design-judge subagent (Task(model: "sonnet")) — runs in parallel with CI. Inputs: diff URL, docs/TASTE.md, docs/ARCHITECTURE.md. Returns findings_count + subagent_tokens.
  3. Mechanical pre-merge gate (orchestrator-inline) — §0 scope, §1 branch, §2 conventional commits, §3 no artifacts, §7 no CHANGELOG edits, §8 secrets grep. Emits one §N name: PASS/BLOCKER line per section.

Read the full file on GitHub · 147 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 · 147 lines · 57 tokens per session scan A 01e2220c7ef0

Subscribe to this mod's changes

auto-iter is a command published in the GitHub repository proggarapsody/bitbottle (3 stars, last pushed 15d ago), licensed MIT. It adds 57 tokens to every session and 2,183 once invoked, about $0.0003 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 2026-08-31.