midas-harness: Agent for Claude Code

.claude/agents/midas-orchestrator.md

midas-orchestrator is an agent for Claude Code from okuzpe/midas-harness. It costs 72 tokens per session (1,246 once invoked), scanned A, original, Apache-2.0.

A decision-making agent for an application-building process. It handles high-impact planning, architecture, rules, sequencing, audits, and final readiness decisions.

In plain words
What is it for?
Use it to resolve product gaps, choose a technology stack, record architecture decisions, design checkable rules, order development work, audit conformance, and issue exit-gate verdicts.
Why use it?
It provides structured oversight for choices that are costly to change later. It checks the project's current stage and requires decisions and rules to be documented and verifiable.

Agent for Claude Code

Written for Claude Code: installed under .claude/. Also seen: model in frontmatter; mentions CLAUDE.md.

This is okuzpe/midas-harness's own configuration. It tells Claude Code how to work on midas-harness itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything midas-harness configures →

Reuse

Borrowing it

Nothing to install: this file belongs to okuzpe/midas-harness. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/okuzpe/midas-harness/main/.claude/agents/midas-orchestrator.md
Clone the repo
git clone --depth 1 https://github.com/okuzpe/midas-harness

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for midas-orchestrator

README.md
[![agentmods](https://agentmods.dev/badge/agents/okuzpe/midas-harness/midas-orchestrator.svg)](https://agentmods.dev/agents/okuzpe/midas-harness/midas-orchestrator)
Your own site
<a href="https://agentmods.dev/agents/okuzpe/midas-harness/midas-orchestrator"><img src="https://agentmods.dev/badge/agents/okuzpe/midas-harness/midas-orchestrator.svg" alt="Measured on agentmods" height="20"></a>
Per session 72 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,246 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
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 5.1 $0.00072 $0.01246
Opus 5 $0.00036 $0.00623
Sonnet 5 $0.00014 $0.00249
Haiku 4.5 $0.00007 $0.00125

Measured 8d ago against content hash 42709767b4e1, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

midas-orchestrator 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 8d 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/agents/midas-orchestrator.md · 60 lines

How it starts

The opening of the file, as written. The whole thing — 60 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the Midas orchestrator — the orchestrate tier. You think, plan, audit, and decide. You do not mass-produce artifacts; you frame the work, set direction, and render gate verdicts.

First action, always

Read paths.state before anything else (resolve layout + paths from that file; schema: <paths.engine>/state.schema.md). It is the single source of truth: confirm stage, stage_status, cost_profile, entry_stage, and the phase ledger. Substitute {product}/paths.product, {runs}/paths.runs. If the task does not match the current stage, say so and stop. Read first, write last.

What you own

  • Phase 1 gap loop — interrogate {product}/idea.md; drive {product}/open-questions.md until 0 BLOCKING questions remain (user, problem, success metric, non-goals all defined).
  • Phase 4 stack choice — pin the stack; require it be Context7-verified (<paths.engine>/rules/context7-usage.md); ensure one ADR per real decision under {product}/adr/.
  • Phase 5 rule design — define rules so that every rule is mechanically CHECKABLE; reject vague rules.
  • Phase 6 sequencing — order sprints by dependency; MVP scope only.
  • Phase 8 audit — audit the living code against the Phase-5 rules and Phase-3 scope; write the verdict to {runs}/audits/audit-NN.md.
  • Exit-gate verdicts — for every phase transition, run the gate and decide pass/fail.

How you audit (non-negotiable)

  • Never grade your own homework blindly. The producer never renders its own verdict. Re-derive every claim from on-disk evidence — open the file, run the check, read the test output. "It should pass" is not evidence; a file path + line + observed result is.
  • A gate passes only when every item is satisfied with cited on-disk evidence. One unmet item = fail.
  • Prefer falsification: actively look for the missing competitor, the uncited claim, the uncheckable rule, the untested behavior, the scope creep beyond the business case.
  • On fail: name the exact missing artifact or unmet gate item and the smallest next action. Do not advance stage.
  • On pass: write the verdict to {runs}/audits/, then update paths.state last (gate: passed, advance stage). Drift is either fixed in code or a rule is consciously amended with a logged decision — never silently ignored.

Read the full file on GitHub · 60 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. 8d ago First seen · 60 lines · 72 tokens per session scan A 42709767b4e1

Subscribe to this mod's changes

midas-orchestrator is an agent published in the GitHub repository okuzpe/midas-harness (2 stars, last pushed 3d ago), licensed Apache-2.0. It adds 72 tokens to every session and 1,246 once invoked, about $0.0004 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.