Borrowing it
Nothing to install: this file belongs to ai-forever/harness-bench-fast. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/ai-forever/harness-bench-fast/main/AGENTS.mdgit clone --depth 1 https://github.com/ai-forever/harness-bench-fastWrote 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/instructions/ai-forever/harness-bench-fast/agents-md)<a href="https://agentmods.dev/instructions/ai-forever/harness-bench-fast/agents-md"><img src="https://agentmods.dev/badge/instructions/ai-forever/harness-bench-fast/agents-md/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/instructions/ai-forever/harness-bench-fast/agents-md"><img src="https://agentmods.dev/badge/instructions/ai-forever/harness-bench-fast/agents-md.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 | $0.01425 | $0.01425 |
| Opus 5.5 | $0.00570 | $0.00570 |
| Sonnet 5.5 | $0.00285 | $0.00285 |
| Haiku 4.5 | $0.00143 | $0.00143 |
Grade A, and why
harness-bench-fast 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 6d 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.
How it starts
The opening of the file, as written. The whole thing — 78 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Repository Instructions
Benchmark Result Tables
When reporting итоговую таблицу benchmark runs for this repository, use this table format by default:
| harness | model | passed | score | steps | tokens |
|---|
Rules:
- Use the canonical public suites by default: frozen Light (391 tasks, IDs 1–391) and Medium (90 tasks: IDs 392–471 plus selected wave 21). Publish an overall suite score only with complete valid coverage. The unchanged 531-task registry/default includes archival controls and is not the public leaderboard denominator. Current task-set membership version is v0.21.3. Preserve historical 80-task Medium scores under their original scope.
- Sort rows by
scoredescending. - Use
passedas<passed>/<suite total>andscoreas a percentage with one decimal place. - Do not mix task-set versions in one table. Earlier runs are not comparable: v0.14.0 scored 371 tasks, v0.16.0 corrected defects across the 391 of v0.15.0 without changing their number, v0.17.0 adds the long-context wave as tasks 392–411, v0.18.0 adds the tool-reflection wave as tasks 412–431, and v0.19.0 adds wave 19 (feasibility) as tasks 432–471. If more than one version must appear, split the table and name the task-set version in each.
- v0.19.1 changes execution/scoring protocol and clarifies task 410 without adding tasks. Preserve historical v0.19.0 scores and protocols; do not relabel them as v0.19.1. Use a fresh result JSON: resume rejects a different task-set version and wave 19 runtime/contract hash.
- Do not include artifact links or artifact paths in the main table unless explicitly requested.
- Use the human-readable harness and model names, not only the JSON filename.
- Include
stepsandtokensfrom the run JSON when present. - Publish token usage only for the final retained attempt of each task, whether passed or failed. Exclude superseded infrastructure/recovery attempts from public tables, protocol pages and aggregates; keep all-attempt accounting private.
- If a run artifact does not contain steps or token metrics, show
0and note that0means the metric is absent from the artifact, not that nothing was spent.
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.
- 6d ago Changed · +11 lines · +238 tokens per session 16bf50083ef4
- 8d ago Changed · +15 lines · +306 tokens per session 1815c6cae8f1
- 11d ago Changed · +4 lines · +60 tokens per session c875c3857cae
- 27d ago Changed · +19 tokens per session 3b47fb59ce86
- 29d ago Changed · -16 tokens per session 3ab697afdb13
- 1mo ago Changed · +18 tokens per session a096bac22318
- 2mo ago First seen · 48 lines · 800 tokens per session scan A 4b658db91926
harness-bench-fast AGENTS.md is an instructions file published in the GitHub repository ai-forever/harness-bench-fast (55 stars, last pushed 2d ago), licensed MIT. It adds 1,425 tokens to every session, about $0.0057 per session on Opus 5.5. 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 2026-08-30.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).
deno copilot-instructions.md
Copilot instructions for denoland/deno, covering deno development guide for github copilot, network access, high level overview, key directories and key files to understand first.