devtool-ax-kit AGENTS.md

Repository instructions for Devtool AX Kit, a toolkit for carefully testing and evaluating AI coding agents. It records experiments, checks results independently, compares outputs, and produces evidence-based reports.

In plain words
What is it for?
Designing controlled tasks, capturing and replaying agent runs, comparing results, validating outcomes, and preparing reports about agent behavior.
Why use it?
They protect test validity and sensitive information by separating verification from agent edits, keeping credentials out of the repository, and distinguishing unavailable resources from real failures.

Instructions file for CodexOpenCode

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 instructions/shenli/devtool-ax-kit/agents-md
Clone the repo
git clone --depth 1 https://github.com/shenli/devtool-ax-kit

Made for: Codex, OpenCode.

Per session 241 This file is loaded in full into every session.
When invoked 241 The same file — it is already loaded in full.
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 $0.00241 $0.00241
Opus 5 $0.00120 $0.00120
Sonnet 5 $0.00048 $0.00048
Haiku 4.5 $0.00024 $0.00024

Measured yesterday against content hash eda90fe8361b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

devtool-ax-kit AGENTS.md 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.md · 31 lines

What it actually says

Agent instructions

Read the task prompt, then consult the relevant skill in skills/:

  • ax-task-design for controlled task and fixture design
  • ax-verification for independent deterministic checks
  • ax-run-capture for trajectories, checkpoints, receipts, and replay
  • ax-comparison for raw-vs-official surface comparisons
  • ax-failure-report for evidence-led vendor reports

The methodology and outcome vocabulary are in docs/agent-ax-methodology.md and docs/outcome-taxonomy.md. These are the repository sources of truth.

Invariants

  • Never put real credentials, tokens, production data, or raw secret-bearing trajectories in the repository.
  • Keep the verifier outside the agent-editable workspace and do not let the agent modify it.
  • Record external writes only with a stable identity and authoritative receipt.
  • Treat unavailable resources separately from functional failures.
  • Do not repeat an uncertain external write; use pinned replay and the original receipt instead.

Validation

python3 -m unittest discover -s tests -v
python3 scripts/replay_experiment.py
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 · 31 lines · 241 tokens per session scan A eda90fe8361b

Subscribe to this mod's changes

devtool-ax-kit AGENTS.md is an instructions file published in the GitHub repository shenli/devtool-ax-kit (2 stars, last pushed 3d ago), licensed Apache-2.0. It adds 241 tokens to every session, about $0.0012 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.