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/cosmtrek/mindwalk/agents-mdgit clone --depth 1 https://github.com/cosmtrek/mindwalkWhat 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 | $0.00876 | $0.00876 |
| Opus 5 | $0.00438 | $0.00438 |
| Sonnet 5 | $0.00175 | $0.00175 |
| Haiku 4.5 | $0.00088 | $0.00088 |
Grade A, and why
mindwalk 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 — 46 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
mindwalk is a local visualizer for coding-agent sessions. It supports Claude Code, Codex, and pi, turning agent session logs plus repository structure into a deterministic 3D "code city" that can be explored in a browser.
Design
The project has three primary artifacts:
- A normalized trace of what happened during a supported coding-agent session.
- A deterministic citymap of the repository being edited or inspected.
- An evaluation report: an LLM judge's findings about one session — four fixed process dimensions plus a task-specific rubric layer — generated on explicit request only.
The UI combines those artifacts so users can see how a coding agent moved through a codebase over time — and, when asked, how well. Keep the separation clear: source-specific parsing should not know about rendering, citymap generation should not depend on session playback, the judge reads only the normalized trace (never raw session logs), and the server should mainly connect data sources to the web client.
Architecture
cmd/mindwalkprovides the CLI commands: serve a local UI, open a session, build a citymap, export a trace, or evaluate a session.internal/adapterconverts supported agent session formats into the shared model. Claude Code, Codex, and pi each have an adapter; keep every source, current and future, behind its adapter boundary.internal/modelowns the trace, citymap, and report data contracts.internal/citymapbuilds deterministic layouts from repository contents.internal/judgerenders a trace into an evidence document and runs a sealed local agent CLI (claude or codex) over it in up to two calls: the first drafts a task rubric — task-grouped criteria derived from the session's user messages — and the second is one unified scoring pass over the four fixed dimensions plus any rubric criteria. The rubric phase can skip (no events, no or too-little task text), reuse the cached report's rubric when the task wording is unchanged, or degrade to a dimensions-only report when generation fails; it never blocks the fixed layer. The judge subprocess gets no tools; verdicts — per dimension and per criterion — are always derived mechanically from finding severities and coverage, never decided by the LLM. Reports are cached in~/.mindwalk/reports;docs/dynamic-rubric-evaluation.mdexplains the rubric layer.internal/serverexposes local APIs and serves the web app.internal/server/staticholds the embedded frontend assets generated fromweb/dist.webcontains the React, Vite, and Three.js frontend.schemamirrors the exported JSON contracts.
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 · 46 lines · 876 tokens per session scan A fa435f02a4aa
mindwalk AGENTS.md is an instructions file published in the GitHub repository cosmtrek/mindwalk (1,305 stars, last pushed 23d ago), licensed MIT. It adds 876 tokens to every session, about $0.0044 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 2026-08-30.
Other instructions, from other repositories
Browser4 CLAUDE.md
Instructions for platonai/Browser4, covering browser4 — project context for claude, architecture, key dispatch chain (cli → browser), batch commands and e2e test structure.
spring-ai-agentcore AGENTS.md
Instructions for spring-ai-community/spring-ai-agentcore, covering agents.md, project overview, architecture, key components and artifact store classes.
flyto-core CLAUDE.md
Instructions for flytohub/flyto-core, covering claude notes, cross-agent handoff and shared code intelligence.
deckforge AGENTS.md
Instructions for tph-kds/deckforge, covering deckforge agent entry point, code intelligence, read order, default routing and non-negotiable implementation rules.
fast-browser CLAUDE.md
Claude Code instructions for m4ttstack/fast-browser, covering fast browser plugin, where a change belongs, fork branch: use fast-browser-runtime, releasing a new runtime and re-pinning this repo: use the script.
agentoscope AGENTS.md
Instructions for rafaelcg/agentoscope, covering agentoscope — agent instructions, what this is, layout, commands and conventions.