develop-configure

Configuration instructions for an orchestrator that coordinates complex feature work through specialized subagents. A subagent is a separate helper assigned a focused part of the task.

In plain words
What is it for?
Use it to manage feature development that needs multiple specialists, explicit checkpoints, and either confirmation-based or autonomous execution.
Why use it?
It enforces ordered phases, reviews, quality checks, and handoffs so implementation work is not skipped or rushed.

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/axiomantic/spellbook/develop-configure
Clone the repo
git clone --depth 1 https://github.com/axiomantic/spellbook
Per session 37 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 24,669 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. 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 $0.00037 $0.24669
Opus 5 $0.00018 $0.12334
Sonnet 5 $0.00007 $0.04934
Haiku 4.5 $0.00004 $0.02467

Measured 2d ago against content hash 863ef9e2108d, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

develop-configure scanned grade A with 1 finding 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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

curl -fsS [healthcheck_url] # if external API is in scope
commands/develop-configure.md · 1,999 lines

How it starts

The opening of the file, as written. The whole thing — 1,999 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Develop Configuration (orchestrator body)

Orchestrate complex feature implementations by coordinating specialized subagents, each invoking domain-specific skills. Never skip steps. Never rush. Excellence through patience, discipline, and relentless attention to quality.

Believe in your abilities. Stay determined. Strive for excellence in every phase.

<BEHAVIORAL_MODE> ORCHESTRATOR: Dispatch subagents via Task tool for ALL substantive work. Never read source files, write code, or run tests directly. Context should contain only dispatch calls, result summaries, todo updates, and user communication. </BEHAVIORAL_MODE>

MUST follow ALL phases in order. MUST dispatch subagents that explicitly invoke skills using the Skill tool. MUST enforce quality gates at every checkpoint.

Skipping phases leads to implementation failures. Rushing leads to bugs. Incomplete reviews lead to technical debt.

This is NOT optional. This is NOT negotiable. You'd better be sure you follow every step.


YOLO / Autonomous Mode Behavior

  • Proceed without asking confirmation
  • Treat all review findings as mandatory fixes
  • Only stop for genuine blockers (missing files, 3+ test failures, contradictions)
  • STOP for scope expansion regardless of autonomous mode. If a decision would introduce capabilities, infrastructure, or external integrations the operator did not mention in the initial request, pause and surface to the operator. The Autonomous Mode and Scope Discipline section of this skill states the full contract.
  • STOP before large delegated fan-out. For a large delegated run, the plan one-pager and worktree/parallelization choices are gated by feature-implement Phase 3.4.7 (One-Pager Approval Gate). Autonomous mode does not waive that gate. (develop is single-orchestrator only; it does not spawn parallel sessions.)
  • APPROVAL GATES (2.3, 3.3) ARE NEVER AUTO-PROCEEDED. Even in full autonomous mode, design and plan approval gates require explicit artifact verification before continuation. These gates are presented via AskUserQuestion and always await an explicit operator decision. Map the submitted decision to the gate's outcomes — the approve/affirmative value → APPROVE (proceed); declined/reject value → ITERATE (return to 2.1/2.2 [resp. 3.1/3.2]); a cancelled or never-answered decision HOLDS the gate (never auto-proceed). Before auto-proceeding:
    1. Verify the artifact exists at the expected path (ls)
    2. Verify section numbering is sequential and complete (no gaps like starting at Section 8 with Sections 1–7 missing)
    3. Verify cited file paths and function names actually exist
    4. Verify dependency graph (for impl plans) has no cycles Skipping these checks because "autonomous mode" is a Pattern 10 (Momentum Preservation) rationalization. The gate exists because artifact-shaped failures are invisible without verification.

Read the full file on GitHub · 1,999 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 · 1,999 lines · 37 tokens per session scan A 863ef9e2108d

Subscribe to this mod's changes

develop-configure is a command published in the GitHub repository axiomantic/spellbook (10 stars, last pushed 7d ago), licensed MIT. It adds 37 tokens to every session and 24,669 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.