calibration-contract

calibration-contract is an agent for coding agents from get-tmonier/argot. It costs 0 tokens per session (2,493 once invoked), scanned A, original, MIT.

A binding technical contract for calibrating a code-scoring system with multiple random seeds and a median threshold. It defines how calibration should avoid hiding improvements that affect both calibration and test examples equally.

In plain words
What is it for?
Use it when changing the scoring or calibration code described in the contract. It guides threshold measurement and the treatment of optional scoring contributions.
Why use it?
A score contribution that appears equally often in calibration and test data can raise the threshold by the same amount, producing little or no improvement in detected issues. The contract addresses that cancellation problem.

Agent

Part of the argot plugin — 7 skills, 5 agents, 1 hook, 1 MCP server shipped together

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/get-tmonier/argot/calibration-contract
Clone the repo
git clone --depth 1 https://github.com/get-tmonier/argot

Or install argot, the plugin that ships this one along with the rest of its 7 skills, 5 agents, 1 hook, 1 MCP server.

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 calibration-contract

README.md
[![agentmods](https://agentmods.dev/badge/agents/get-tmonier/argot/calibration-contract.svg)](https://agentmods.dev/agents/get-tmonier/argot/calibration-contract)
Your own site
<a href="https://agentmods.dev/agents/get-tmonier/argot/calibration-contract"><img src="https://agentmods.dev/badge/agents/get-tmonier/argot/calibration-contract.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,493 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 $0.00000 $0.02493
Opus 5 $0.00000 $0.01247
Sonnet 5 $0.00000 $0.00499
Haiku 4.5 $0.00000 $0.00249

Measured 5d ago against content hash 97e3023e09d2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

calibration-contract 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 5d 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.

docs/agents/calibration-contract.md · 230 lines

How it starts

The opening of the file, as written. The whole thing — 230 lines — stays where its author put it; the contents beside it link to each section on GitHub.

G7 Calibration Contract

Status: binding (era-13.5 gate G7) Code: crates/argot-core/src/scoring/calibration.rsrun_calibrate (multi-seed median threshold)


The problem: symmetric firing cancels recall

During era-13 Phase 10 development, every new optional contribution tested (cluster-rare-threshold rule, Phase-4 ShapePrimitive penalties) was measured against the calibrated threshold and found to add ~zero recall. The cause is structural:

Any additive contribution that fires symmetrically on calibration hunks and fixture hunks inflates the per-corpus threshold by the same magnitude it adds to fixture scores. The net catch impact is ~zero.

Empirical evidence: the era-13 Phase A scout (docs/research/evidence/era13-final.md § cancellation) confirmed that with cluster_rare_threshold=2, all three corpora (faker-js, fastapi, hono) had exactly 200–280 rare-branch fires per calibration seed. The threshold rose by exactly cluster_bonus = 5.0 — matching the observed scout t_sym − t_asym = 5.0000 on every corpus.


Why suppressing on the cal path is mathematically sound

The cancellation argument assumes both paths use the same scorer. The fix is to break that symmetry at the calibration boundary, not inside the scorer.

Era-11 cluster_bonus is already asymmetric by construction

Calibration hunks are sampled from model_a_files — files whose callee bags are typical for their cluster. By construction, these files' callees are within their cluster's attested set, so the cluster-absent-callee branch of weighted_contribution_for_file does not fire on calibration hunks. Era-11's cluster_bonus is asymmetric without any special plumbing.

New optional contributions are not asymmetric by construction

The cluster-rare rule fires when a callee is attested in ≤ N cluster files — a Zipf-distribution property shared by typical code and anomalous code alike. Calibration hunks from model_a_files call rare-attested functions too (e.g. build utilities, test helpers that appear in one file). Shape primitives based on AST patterns (exception handling ratios, call scope fractions) also fire on calibration hunks unless the corpus is carefully curated.

Read the full file on GitHub · 230 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. 5d ago First seen · 230 lines · 0 tokens per session scan A 97e3023e09d2

Subscribe to this mod's changes

calibration-contract is an agent published in the GitHub repository get-tmonier/argot (48 stars, last pushed 2d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,493 tokens. 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.