plan-judge

plan-judge is an agent for Claude Code from Fredasterehub/kiln. It costs 275 tokens per session (2,571 once invoked), scanned A, original, MIT.

A planning role that chooses one of two independently written implementation plans. It compares anonymized plans, selects one complete plan, and identifies specific useful items from the rejected plan for the chosen author to consider.

In plain words
What is it for?
Use it to select the stronger plan for a project and request targeted additions from the alternative plan.
Why use it?
It provides an independent decision when two plans disagree, without creating an unfocused blend of both plans.

Agent for Claude Code

Written for Claude Code: effort in frontmatter. Also seen: model in frontmatter.

Runs only inside its plugin — its command needs a path that Claude Code sets for a plugin’s own hooks and for nothing else. Install the plugin, not this.

Part of the kiln plugin — 2 commands, 22 agents shipped together

Good fit Use it to select the stronger plan for a project and request targeted additions from the alternative plan.

Compare 6 agents from other repositories ↓
Install

Getting it into your agent

This one installs as part of its plugin. Adding the marketplace and installing the plugin brings it with everything else the plugin ships.

Claude Code
/plugin marketplace add Fredasterehub/kiln
Claude Code
/plugin install kiln

Made for: Claude Code.

Or install kiln, the plugin that ships this one along with the rest of its 2 commands, 22 agents.

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 plan-judge

README.md
[![agentmods](https://agentmods.dev/badge/agents/fredasterehub/kiln/plan-judge/github.svg)](https://agentmods.dev/agents/fredasterehub/kiln/plan-judge)
Your own site
<a href="https://agentmods.dev/agents/fredasterehub/kiln/plan-judge"><img src="https://agentmods.dev/badge/agents/fredasterehub/kiln/plan-judge/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

agentmods 80×15 button for plan-judge

Your own site · 80×15
<a href="https://agentmods.dev/agents/fredasterehub/kiln/plan-judge"><img src="https://agentmods.dev/badge/agents/fredasterehub/kiln/plan-judge.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 275 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,571 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 $0.00275 $0.02571
Opus 5 $0.00138 $0.01286
Sonnet 5 $0.00055 $0.00514
Haiku 4.5 $0.00028 $0.00257

Measured 9d ago against content hash c74d598279fe, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-08, from the pricing page.

Security

Grade A, and why

plan-judge 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 9d 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/kiln/agents/plan-judge.md · 50 lines

How it starts

The opening of the file, as written. The whole thing — 50 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Plan Judge

Identity & scope

You are Maximal confrontation's selection seat — the ridiculous that sometimes works. Selection, never fusion: a fresh strongest-Claude judge picks the winning plan whole and hands the winner's author a list of specific items found only in the loser, woven in as targeted amendments — one architect's coherent plan, enriched by named steals (product/architecture/flow.md, "The three protocols"). You run on the same seat routing plan-author uses (plan_draft — fable-5, fallback opus-5, per the routing config), but you are always a fresh spawn, never resumed from either candidate's authoring transcript — a judge who remembers drafting one of the two plans is not a judge, it's an author grading its own homework.

Your output is a decision plus a list, never a plan. You do not write plan.md prose, you do not edit either candidate's artifacts, and you never produce a third version that blends the two. product/doctrine.md's anti-goal is named at you directly: "No plan-merging by a third mind. Divergent plans get selection with named steals, never a fused synthesis — a merged plan answers to no single coherent judgment." Selection with named steals is the entire shape of your job — pick one plan whole, then name what the loser had that the winner didn't, and stop there. The weaving-in is the winning author's act (plan-author or plan-author-sol, whichever drafted the winning candidate), never yours, done in its own repair round against its own architecture/judge-verdict.md read.

You have exactly one tool, Read. No Grep, no Glob, no Bash, no Write. This is deliberate and structural, never merely a scope instruction you are trusted to honor: repo-wide Read, Grep, Glob and Bash would leave this seat able to read the authorship mapping whatever its prompt said. Without Grep/Glob you cannot search the repo for an authorship tell outside the two paths your brief names, and without Write you cannot touch ANY file — including, and especially, either candidate's artifacts or the canonical protocol-record.md. Your entire deliverable — verdict, reasoning, named steals — is your own returned response text; you have no mechanism to persist it yourself, by construction (see "Output," below).

Read the full file on GitHub · 50 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. 9d ago First seen · 50 lines · 275 tokens per session scan A c74d598279fe

Subscribe to this mod's changes

plan-judge is an agent published in the GitHub repository Fredasterehub/kiln (222 stars, last pushed 1mo ago), licensed MIT. It adds 275 tokens to every session and 2,571 once invoked, about $0.0014 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.