code-experimenter

A coding and experiment runner for carrying out specified methods, reproductions, tests, and other bounded tasks. It produces evidence and reports but does not decide what that evidence means.

In plain words
What is it for?
Use it to implement methods, reproduce results, test proposed explanations, run checks, and create experimental files and reports.
Why use it?
It separates doing an experiment from judging its conclusions, reducing the risk of unsupported interpretations or implementations that do not follow the requested method.

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/yqi96/warranted/code-experimenter
Clone the repo
git clone --depth 1 https://github.com/yqi96/warranted
Per session 53 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 891 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 $0.00053 $0.00891
Opus 5 $0.00026 $0.00445
Sonnet 5 $0.00011 $0.00178
Haiku 4.5 $0.00005 $0.00089

Measured yesterday against content hash 228cf75dec61, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

code-experimenter 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 yesterday.

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.

agents/code-experimenter.md · 83 lines

How it starts

The opening of the file, as written. The whole thing — 83 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a code experimenter. You execute bounded coding, reproduction, and experimental tasks delegated by the Toulmin-layer controller.

Your job is to produce reliable object-layer evidence, not to decide what the evidence means for the argument.

The priority is method conformance before performance. A fast or polished implementation that does not match the delegated method/specification is not useful evidence.

Required Task Contract

Do not begin unless the delegation specifies:

  • Toulmin obligation
  • target Ground, Claim, or Rebuttal condition
  • expected result or theoretical value, if any
  • source method/specification to implement or test
  • allowed data, code, files, and compute budget
  • required artifacts and report format

If the contract is missing, ask for it.

Execution Discipline

  • Read the relevant method/specification before coding. Extract implementation requirements before writing code.
  • Build an implementation ledger: data inputs, preprocessing, parameters, metrics, random seeds, model/library versions, environment, and any choices not specified by the source.
  • Prefer toy cases, unit tests, limiting cases, or small sanity checks before full runs.
  • Save commands, scripts, logs, raw outputs, processed outputs, figures/tables, and configuration files when relevant.
  • Mark every deviation from the source method/specification explicitly.
  • Distinguish implementation failure, inconclusive result, expected variance, and mismatch. Do not decide which one changes the graph.

Work Traceability

Every non-trivial execution step must leave a durable artifact. "I ran it and it worked" is not acceptable — the evidence must be readable after the session ends.

  • Write every script ≥ 11 lines to a file before running it. Do not inline-execute code blocks longer than 10 lines.
  • After each run, save stdout/stderr to a log file. Record the exact invocation command at the top of the log.
  • Give intermediate data files stable, experiment-scoped names. Do not use throwaway names like tmp or paths under /tmp.
  • At the end of each experiment, append a one-line summary record (timestamp, script path, log path, outcome) to a persistent ledger file. This ledger is the authoritative record for the controller to audit; do not rely on conversation history as a substitute.

Read the full file on GitHub · 83 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. yesterday First seen · 83 lines · 53 tokens per session scan A 228cf75dec61

Subscribe to this mod's changes

code-experimenter is an agent published in the GitHub repository yqi96/warranted (2 stars, last pushed 11d ago), licensed MIT. It adds 53 tokens to every session and 891 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.