harness-doctor

A read-only repository check that measures how ready a project is for reliable collaboration with AI coding agents. It gives a score based on instructions, limits, feedback, memory, evaluation, and governance.

In plain words
What is it for?
Use it to inspect repository guidance, checks, memory records, evaluation methods, and governance without changing files, then produce a Harness Score and diagnostic report.
Why use it?
It reveals missing or disconnected project practices that may cause agents to repeat mistakes across sessions.

Command

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 commands/harnessworks/harness-starter-kit/harness-doctor
Clone the repo
git clone --depth 1 https://github.com/harnessworks/harness-starter-kit
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 1,324 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 $0.01324
Opus 5 $0.00000 $0.00662
Sonnet 5 $0.00000 $0.00265
Haiku 4.5 $0.00000 $0.00132

Measured yesterday against content hash 342a5276ce94, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

harness-doctor 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.

commands/harness-doctor.md · 166 lines

How it starts

The opening of the file, as written. The whole thing — 166 lines — stays where its author put it; the contents beside it link to each section on GitHub.

/harness doctor

Run Harness Doctor to evaluate how ready the current repository is for reliable AI coding agent collaboration.

Harness Doctor is diagnostic only. Do not create, edit, delete, move, format, or stage files while running this command.

Do not remove a nested ./harness-starter-kit directory from the target repository while running this command. If you created a temporary clone outside the target repository, such as under /tmp, you may remove only that temporary copy after the report is complete.

Goal

Inspect the repository and produce a Harness Score out of 100. The score should identify weak points where AI coding agents are likely to repeat mistakes across sessions. It is a harness health diagnostic, not proof of agent effectiveness or mistake reduction.

Scope

Harness Doctor scores the six-part repository harness model in docs/theory/harness-engineering.md:

Harness = Instructions + Constraints + Feedback + Memory + Evaluation + Governance

The score is a harness health and coupling diagnostic. It checks durable repository evidence for the six elements and reports broken links between rules, checks, memory, evaluation, and governance. It does not prove agent effectiveness or mistake reduction.

Doctor may report task outcome records and effectiveness reports as evidence that the Evaluation element exists. It must not count those records as proof that agents improved unless the report explicitly includes comparable outcome evidence. Runtime-only execution sandboxes and tool protocols are outside the score unless their assumptions are visible in repository files, commands, CI, or approval rules.

Use this principle when judging the repository:

Prompting is temporary. Context is session-scoped. A harness is project-scoped.

Procedure

  1. Treat the current working directory as the target repository root.
  2. Inspect durable repository files, not chat context.
  3. If ./harness-starter-kit exists, treat it as read-only reference material. Do not delete it, ignore it, move it, or clean it up as part of the doctor run.
  4. Read likely harness files:
    • AGENTS.md, CLAUDE.md, .cursor/rules/*, .github/copilot-instructions.md
    • README.md, CONTRIBUTING.md, project docs, and CI configs
    • docs/decisions, docs/failures, docs/conventions, and docs/domain
    • scripts/, package manager config, test config, lint config, and pre-commit config
  5. If available, run a baseline scan:

Read the full file on GitHub · 166 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 · 166 lines · 0 tokens per session scan A 342a5276ce94

Subscribe to this mod's changes

harness-doctor is a command published in the GitHub repository harnessworks/harness-starter-kit (113 stars, last pushed 2mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,324 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.