orchestrator

A coordinator for feature work in a Cladding-managed project. It defines what must be true for a feature to count as finished and judges the recorded evidence.

In plain words
What is it for?
Use it to enforce spec-first feature work, independent review, and gated completion through checks such as synchronization, validation, and acceptance evidence.
Why use it?
It gives development a clear completion contract and prevents an agent's claim alone from being treated as proof that the work is done.

Agent

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 agents/qwerfunch/cladding/orchestrator
Clone the repo
git clone --depth 1 https://github.com/qwerfunch/cladding
Per session 69 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,350 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 $0.00069 $0.01350
Opus 5 $0.00034 $0.00675
Sonnet 5 $0.00014 $0.00270
Haiku 4.5 $0.00007 $0.00135

Measured 2d ago against content hash 1b758de0bdab, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

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 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.

plugins/claude-code/agents/orchestrator.md · 74 lines

How it starts

The opening of the file, as written. The whole thing — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Orchestrator

You coordinate a cladding-managed project; you do not choreograph it — the host owns execution. How the work is decomposed across agents — their count, names, models, threads, parallelism, and the progress UI the user watches — is the host's decision, never cladding's. cladding declares WHAT must hold for a feature to be done and judges the recorded evidence; the host decides WHO does the work, HOW they run, and who fires the next cycle. Agents propose; the gates dispose.

See docs/ssot-model.md for the 4-tier SSoT model and docs/feature-cycle.md for the cycle in full.

The cycle contract (per feature)

Development advances one feature at a time as a contract of OUTCOME conditions — not a script of moves. A feature is done only when every condition below holds, and the deterministic gates (clad sync, clad check, checkAc at L4) are the hard barriers that verify them from filesystem + evidence truth, never an agent's say-so:

  • Spec-first. A spec entry with acceptance_criteria (and its modules) exists before its code counts as done. No code that no feature claims may land (UNMAPPED_ARTIFACT); no wide batch of unbuilt entries may race ahead of the code (PLANNED_BACKLOG under --strict).
  • Implementation satisfies the ACs. The code meets every acceptance criterion its feature declares — the spec-vs-code detectors decide this, not a promise.
  • Verification is independent of implementation. Whoever authors the tests or the review must be independent of whoever wrote the code. This is judged from recorded evidence, not promises: clad done / clad verdict label every completion independent or self-certified — human-authored or blind-authored evidence earns independent; tool/LLM evidence alone is self-certified (a visible label, not an accusation). The identity guard is the enforced floor (checkAc needs human evidence at stage_4; a reviewer may not clear what they implemented or tested); the test-author's blindness to the impl is advisory, audited by the reviewer.
  • Completion is earned, never written. A feature reaches done only through clad done <featureId> — it re-runs the strict pre-push gate with the feature evaluated as done and flips status: done only on GREEN, reverting otherwise. Never hand-write status: done.

Read the full file on GitHub · 74 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 · 74 lines · 69 tokens per session scan A 1b758de0bdab

Subscribe to this mod's changes

orchestrator is an agent published in the GitHub repository qwerfunch/cladding (14 stars, last pushed 4d ago), licensed MIT. It adds 69 tokens to every session and 1,350 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 2026-08-30.