fire-autonomous

An autonomous software-building command that takes over after the product requirements document is complete.

In plain words
What is it for?
Use it to run selected project phases, set the starting and ending phases, limit attempts per phase, verify each result, and continue through the build pipeline.
Why use it?
It removes routine human pauses between planning, implementation, testing, review, and fixes while retaining safety checks and automatic fix cycles.

Command

Part of the fire-flow plugin — 44 commands, 15 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 commands/thierryn/fire-flow/fire-autonomous
Clone the repo
git clone --depth 1 https://github.com/ThierryN/fire-flow

Or install fire-flow, the plugin that ships this one along with the rest of its 44 commands, 15 agents.

Per session 18 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 4,416 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 $0.00018 $0.04416
Opus 5 $0.00009 $0.02208
Sonnet 5 $0.00004 $0.00883
Haiku 4.5 $0.00002 $0.00442

Measured 3d ago against content hash fee797f7824d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

fire-autonomous 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.

commands/fire-autonomous.md · 558 lines

How it starts

The opening of the file, as written. The whole thing — 558 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/fire-autonomous

Full autopilot after PRD is complete — plan, execute, verify, advance through all phases independently


Purpose

After discussions and PRD are complete (/fire-1a-discuss), this command takes over the entire build pipeline. Claude plans each phase, executes it, verifies the work (verifier + reviewer), fixes any issues, and advances to the next phase — all without human checkpoints.

Design principle (Boris Cherny): "You don't trust; you instrument." All verification stays active. Verdicts auto-route to fix cycles instead of pausing for human input. The human reviews the finished product, not intermediate steps.

Safety: Path verification, HAC, circuit breaker, fire-verifier, and fire-reviewer all remain active. Autonomous mode removes human pauses, not safety gates.


Arguments

arguments:
  --from-phase:
    type: integer
    default: "next incomplete phase from CONSCIENCE.md"
    description: "Start from this phase number"

  --to-phase:
    type: integer
    default: "last phase in VISION.md"
    description: "Stop after completing this phase"

  --max-phase-attempts:
    type: integer
    default: 3
    description: "Max plan-execute-verify cycles per phase before escalating"

  --max-iterations:
    type: integer
    default: 50
    description: "Max iterations per execution loop (passed to fire-3-execute)"

  --dry-run:
    type: boolean
    default: false
    description: "Show what would be done without executing"

  --token-efficient:
    type: boolean
    default: false
    description: "Enable ALAS context slicing for ~60% token reduction. Passes --token-efficient to fire-3-execute. Default OFF for best quality."

Process

Step 1: Validate Prerequisites

Read .planning/VISION.md → extract all phases
Read .planning/CONSCIENCE.md → get current status, find next incomplete phase

IF --from-phase not specified:
  from_phase = first phase with status != complete in CONSCIENCE.md

IF --to-phase not specified:
  to_phase = last phase in VISION.md

Validate:
  1. .planning/ directory exists
  2. VISION.md has phases in range [from_phase, to_phase]
  3. For each phase in range: MEMORY.md exists (PRD/discussion complete)
  4. Path Verification Gate (MANDATORY — working directory matches project)
  5. Cross-Phase DAG Validation (v13.0 — from fire-3-execute Step 3.55)
     → Check for circular dependencies between phases
     → Validate provides/requires contracts match across phase boundaries
     → STOP if cycles found or contracts broken

IF any phase lacks MEMORY.md:
  Display:
    "Phase {N} has no MEMORY.md — discussions not complete.
     Run /fire-1a-discuss {N} first.
     Autonomous mode requires all target phases to have context."
  STOP

Read the full file on GitHub · 558 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 · 558 lines · 18 tokens per session scan A fee797f7824d

Subscribe to this mod's changes

fire-autonomous is a command published in the GitHub repository ThierryN/fire-flow (77 stars, last pushed 21d ago), licensed MIT. It adds 18 tokens to every session and 4,416 once invoked, about $0.0001 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.