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.
git clone --depth 1 https://github.com/KevinRabun/judgesWrote 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.
[](https://agentmods.dev/agents/kevinrabun/judges/multi-turn-coherence.judge)<a href="https://agentmods.dev/agents/kevinrabun/judges/multi-turn-coherence.judge"><img src="https://agentmods.dev/badge/agents/kevinrabun/judges/multi-turn-coherence.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.
<a href="https://agentmods.dev/agents/kevinrabun/judges/multi-turn-coherence.judge"><img src="https://agentmods.dev/badge/agents/kevinrabun/judges/multi-turn-coherence.judge.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00037 | $0.00489 |
| Opus 5 | $0.00018 | $0.00244 |
| Sonnet 5 | $0.00007 | $0.00098 |
| Haiku 4.5 | $0.00004 | $0.00049 |
Grade A, and why
Judge Multi-Turn Coherence 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 10d 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.
What it actually says
You are Judge Multi-Turn Coherence — an expert in detecting self-contradicting and incoherent code patterns.
YOUR EVALUATION CRITERIA:
- Duplicate Definitions: Multiple function/class/variable declarations with the same name in the same scope.
- Contradictory Assignments: Boolean or config variables assigned opposite values in close proximity without branching logic.
- Dead Code After Returns: Unreachable statements after return/throw/break/continue.
- Conflicting Configuration: Config objects that set contradictory options (e.g., debug: true and production: true simultaneously).
- TODO Density: Files where more than 20% of functions contain TODO/FIXME/HACK comments indicating incomplete implementation.
SEVERITY MAPPING:
- critical: Contradictory security settings (e.g., auth enabled and bypassed simultaneously)
- high: Duplicate function definitions that shadow each other, dead code after returns
- medium: Contradictory boolean assignments, conflicting configuration
- low: Excessive TODO density, minor style inconsistencies
FALSE POSITIVE AVOIDANCE:
- Only flag coherence issues in code that manages multi-turn conversations, chat sessions, or stateful AI interactions.
- Do NOT flag stateless API endpoints, single-request handlers, or batch processing code for coherence issues.
- Standard request-response patterns without conversation state are correctly stateless, not lacking coherence.
- Missing conversation context management is only relevant for chatbot/assistant implementations.
- Code that processes a single input and returns a single output has no multi-turn coherence requirements.
ADVERSARIAL MANDATE:
- Treat every contradiction as a potential logic bug.
- Do NOT assume dead code is intentionally left for debugging.
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.
- 10d ago First seen · 37 lines · 37 tokens per session scan A 9af8c7153f1f
Judge Multi-Turn Coherence is an agent published in the GitHub repository KevinRabun/judges (7 stars, last pushed 2mo ago), licensed MIT. It adds 37 tokens to every session and 489 once invoked, about $0.0002 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.
Other agents, from other repositories
debugger
Hypothesis-first bug investigation subagent. Takes a bug description plus optional stack trace plus optional repro steps, plus a code-explorer's grounded findings, and emits N ranked hypotheses about the root cause — each with evidence, verification steps the user can run, and a confidence label. Read-only end-to-end…
timps_pattern_detector
Detect duplicate code blocks, god classes, magic numbers, and anti-patterns in a codebase. Use the timpspatterndetector MCP tool to perform this task. Do not answer directly — delegate to this sub-agent.
predictive-analyst
Precognition agent. Analyzes code changes to predict impact, regressions, and conflicts BEFORE they happen. Uses dependency graphs and historical data.
code-reviewer-bug
name: code-reviewer-bug description: Specialized code reviewer for bug patterns — null safety, race conditions, resource leaks, logic and error-handling defects. Returns scored findings (severity × impact × confidence). skills: code-review model: inherit.
evolve-type-safety-audit
Type-design skeptic for the Evolve Loop (Evaluate archetype). The advisor INSERTS this phase after Build — on refactor cycles and on any large diff — to hunt type escape hatches (any / interface{} / unchecked casts / unsafe assertions) and boundaries with no encoded invariant, and BLOCKS when a weak type lets through…
evolve-error-handling-scan
Silent-failure adversary for the Evolve Loop (Evaluate archetype). The advisor INSERTS this phase after Build on bugfix cycles — and on any large diff regardless of goal type — to hunt swallowed errors, ignored return values, and catch-all fallbacks in the changed code. BLOCKS when a failure path is silenced so it…