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.
npx agentmods add instructions/soliestre/twdd/agents-mdgit clone --depth 1 https://github.com/SoliEstre/twddWhat 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 | $0.00587 | $0.00587 |
| Opus 5 | $0.00293 | $0.00293 |
| Sonnet 5 | $0.00117 | $0.00117 |
| Haiku 4.5 | $0.00059 | $0.00059 |
Grade A, and why
twdd 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.
How it starts
The opening of the file, as written. The whole thing — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Public repository rules
This repository is the public surface of Test-Window-Driven Development (TWDD). Keep every change safe to publish.
Read before changing files
- Read
README.md,docs/status.md, and the document you will change. - Use English for the main document. Update the matching Korean document when meaning changes.
- Treat
ai/twdd-spec.mdas a dense expert and machine-readable projection, not as permission to invent missing decisions. - Treat
skills/twdd/SKILL.mdas the canonical agent workflow. Keep Codex, Claude Code, Gemini CLI, and human documentation aligned with it.
Public boundary
- Never add secrets, personal data, private links, internal reports, agent transcripts, local paths, or private case details.
- Separate verified facts, inferences, research proposals, and unresolved questions.
- Describe TWDD as a research-stage proposal. Do not claim novelty or improvements in speed, quality, or safety without public comparative evidence.
- Preserve the distinction between project-owned MIT-licensed material and linked third-party material.
Writing level
- The introduction site uses L1.1.1: everyday language, short sentences, and explanations before technical terms.
- General documents use L1.1.1 through L1.2.2: plain language with a small amount of defined terminology.
ai/uses L5.5.5: compact expert terminology, explicit invariants, formal states, provenance, and failure conditions.
Method rules
- Safety, money, authorization, data integrity, and legal boundaries always outrank window completion.
- Classify expected failures as
Abort,Isolate,Degrade, orContinuebefore a live window. - Treat unclassified failures as
Abort. - Never hide, swallow, or count a recovered exception as a passing scenario.
- Turn live incidents into fixtures, simulators, or record/replay assets whenever feasible.
- Keep TWDD complementary to IDD, TDD, contract-first work, and domain safety controls.
- Keep the skill agent-neutral. Put runtime-specific metadata in
.codex-plugin/,.claude-plugin/,gemini-extension.json, orskills/twdd/agents/rather than forking the workflow.
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.
- yesterday First seen · 43 lines · 587 tokens per session scan A fb80a12a97f9
twdd AGENTS.md is an instructions file published in the GitHub repository SoliEstre/twdd (1 stars, last pushed 24d ago), licensed MIT. It adds 587 tokens to every session, about $0.0029 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.
Other instructions, from other repositories
supergraph AGENTS.md
Instructions for datit309/supergraph, covering supergraph — mandatory workflows, skills, auto language detection, tiered workflow — pick the right tier first and full pipeline (tier 3).
deploygate-agent-plugin CLAUDE.md
Instructions for DeployGate/deploygate-agent-plugin, covering claude.md, project overview, commands, architecture and testing.
project-legibility AGENTS.md
Instructions for perhapsspy/project-legibility, covering agents.md and 검증.
foreman CLAUDE.md
Instructions for fatihguner/foreman, covering foreman — project context, architecture overview, system layers, core content layers and orchestration layers.
lianhuanhua-skills AGENTS.md
Instructions for littlewindy123/lianhuanhua-skills, covering agents.md, project goal, architecture rules, commands and before committing.
ai-plugin GEMINI.md
Gemini CLI instructions for PostHog/ai-plugin, covering posthog extension, available tool categories and guidelines.