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/klimentij/klimkit/agents-mdgit clone --depth 1 https://github.com/klimentij/klimkitWhat 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 | $0.00495 | $0.00495 |
| Opus 5 | $0.00247 | $0.00247 |
| Sonnet 5 | $0.00099 | $0.00099 |
| Haiku 4.5 | $0.00049 | $0.00049 |
Grade A, and why
klimkit 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 2d 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 — 40 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Repository-local instructions for Klimkit. These supplement the shared home-level agent instructions.
Work log (docs/work)
Substantial human+agent work is journaled live in docs/work/<NNN-DDMMYY-slug>/ (see
docs/work/README.md). While working on such a piece: keep the
current phase folder's LOG.md updated as beats happen (timestamp, one-liner, who drove,
file links), preserve every human message verbatim in exactly one fitting markdown of
the work folder, and open a new NNN-DDMMYY-slug/ phase folder when the work enters a new
logical iteration. Artifacts always live inside a phase folder — never directly in a work
folder (two folder layers under docs/work/, always). Never bulk-load docs/work/ into
context — read a LOG.md, descend selectively.
There are no separate memory/log/reflection state files: action history lives in the
LOG.md files, reflections and durable decisions are numbered notes inside the fitting
work phase, and preferences or process rules that must bind every future session graduate
into this AGENTS.md.
HTML reports and explainers — hosted, auth-protected
Any HTML meant for the human driver to view — explainers, proof reports, dashboards,
anything HTML in docs/work/ — is deployed by default through the harness's native
hosted surface, not just left as a file:
- Claude Code: publish with the built-in Artifact tool (a claude.ai artifact — authenticated, private to the author by default; republishing updates the same URL).
- Codex: deploy with Codex Sites (OpenAI-hosted, workspace-authenticated, private by default; save a version, then deploy).
Always keep the default authenticated visibility — private or workspace-only. Never
enable public link sharing unless Klim explicitly asks. Record the deployed URL in the
phase LOG.md next to the file link. The git-tracked HTML in docs/work/ remains the
source of truth; the hosted page is the viewing surface.
Main Release Reminder
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.
- 2d ago First seen · 40 lines · 495 tokens per session scan A ba19940f3280
klimkit AGENTS.md is an instructions file published in the GitHub repository klimentij/klimkit (7 stars, last pushed 23d ago), licensed MIT. It adds 495 tokens to every session, about $0.0025 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
aiworkspace AGENTS.md
Instructions for a-tokyo/aiworkspace, covering ai workspace, structure, key commands, conventions and code quality.
alcove AGENTS.md
AGENTS.md instructions for OctopusGarage/alcove, covering agents.md and smoke / eval shortcuts.
zaungast AGENTS.md
AGENTS.md instructions for mbe24/zaungast, covering agent policy, commit messages, docs sync, validation — post-change and typescript (primary gate; mirrors ci).
agent-skills-harness AGENTS.md
Instructions for a-tokyo/agent-skills-harness, covering agents.md, project overview, repository structure, skills and setup.
SoDam-Agentic-Eng CLAUDE.md
Claude Code instructions for sodam-ai/SoDam-Agentic-Eng, covering sodamagentic — claude code 개발 지침, 빠른 참조 and 절대 원칙.
skills CLAUDE.md
Claude Code instructions for computerlovetech/skills, a project described as: Agent skills published by computerlove.tech <3.