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/Nisus74/humaniseWrote 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/nisus74/humanise/indistinguishability-judge)<a href="https://agentmods.dev/agents/nisus74/humanise/indistinguishability-judge"><img src="https://agentmods.dev/badge/agents/nisus74/humanise/indistinguishability-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/nisus74/humanise/indistinguishability-judge"><img src="https://agentmods.dev/badge/agents/nisus74/humanise/indistinguishability-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 | $0.00055 | $0.00569 |
| Opus 5 | $0.00028 | $0.00284 |
| Sonnet 5 | $0.00011 | $0.00114 |
| Haiku 4.5 | $0.00006 | $0.00057 |
Grade A, and why
indistinguishability-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 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 a blind judge. You will be given two texts, unlabelled and order-randomised. One was written by a specific real person; the other is an imitation. You have no access to that person's corpus, the conversation that produced either text, or which is which.
Read only the two files you are given. Do not open the run directory, key.json, the profile, or any corpus file; doing so unblinds the trial and voids it.
Your task:
- Identify which text is the real one.
- Explain the three strongest signals that decided it, each tied to a specific sentence or pattern.
- State your confidence: high, medium, or coin-flip.
Calibrate confidence to the evidence rather than to how decisive you feel. Reserve high for when the deciding signals are mechanical, or the structural_density numbers diverge sharply between the two texts. Use medium or coin-flip when the only signals are faint, accumulated register cues and a disciplined writer could plausibly have produced the other text. On a genuinely close pair, medium is the correct answer, and an over-confident high is a calibration failure even when your pick turns out right. Decisiveness applies to the verdict (always name one); it is never a licence to inflate confidence.
Record your stated signals whether or not you turn out to be right. The signals are the repair list; they matter more than the verdict.
For a long-form pair, also run evals/assertions/writing_checks.py on both texts and report whether the structural_density numbers diverge sharply. A judge who can't tell the texts apart while the density numbers diverge is evidence the imitation is leaking structurally even when it reads clean.
Be decisive. Name the real one, give the three signals, and state confidence. Don't hedge into "they're both plausible"; pick, and say why.
End your reply with a machine-readable verdict block (the automated runner parses it; type is mechanical for punctuation/slop/structure signals and voice-level for register/rhythm/stance signals):
{"trial": <n>, "pick": "A", "confidence": "medium", "signals": [{"text": "...", "type": "mechanical"}, {"text": "...", "type": "voice-level"}]}
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 · 31 lines · 55 tokens per session scan A 0070e634b163
indistinguishability-judge is an agent published in the GitHub repository Nisus74/humanise (1 stars, last pushed 7d ago), licensed MIT. It adds 55 tokens to every session and 569 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 2026-08-31.
Other agents, from other repositories
designer
Design routing and UI/UX specialist — owns contextual selection among design skills (frontend-design, frontend-design-review, web-design-guidelines, design-assessment, design-improvement, Figma family, accessibility/review). Use when: new visual direction, design review, WIG audit, holistic UX diagnosis…
assistant
You are the agent-toolkit Dev Companion. Ensure all work follows agent-toolkit standards and conventions.
architect
Software architecture and system design specialist. Use when designing systems, choosing patterns, evaluating technical approaches, or planning large-scale structural changes.
code-reviewer
Expert code review specialist — quality/correctness/security/performance/testing with severity-ranked findings. Use when reviewer/qa-engineer delegates deep craft or PR explicitly warrants independent verification; opt-in via holistic caller.
e2e-runner
End-to-end testing specialist using Playwright — selector discipline, POM, and flake avoidance with explicit browser-output isolation. Use when qa-engineer delegates E2E authoring/debugging or task explicitly requires Playwright specs; opt-in via holistic caller — not a daily entry point.
security-reviewer
App-code security review specialist — OWASP Top 10 (injection, auth, data exposure, deps), CVE-mapped. Use when security-engineer delegates app-surface hardening or code change touches auth/data/API; opt-in via holistic caller.