wayfinder-maps

作者 rengwu已验证

Planetary visualization tool for /wayfinder maps

93
Stars
2
Forks
Go
语言
2026/8/24
添加时间

⚠️ 第三方软件声明

本 Skill 为第三方开源软件,独立托管于 GitHub。SkillTip 仅为信息目录,不控制或维护底层仓库。所显示的安全检查为自动化且范围有限,安装前请自行审查源码。

阅读服务条款

安装

添加到你的 Claude Code skills 目录:

# Add to your Claude Code skills
git clone https://github.com/rengwu/wayfinder-maps

快速入门

使用 wayfinder-maps 等 Skills 的指南。

安全报告

已验证

上次扫描:—

{
  "status": "PASSED",
  "issues": []
}

README.md

demo app: one wayfinder map at a moment

wayfinder-maps

A read-only CLI and viewer for wayfinder maps — the markdown planning memory an agent leaves under .plan/<effort>/ as it charts a large effort as a graph of investigation tickets. The wayfinder method and its skills were created by Matt Pocock; this repo adapts them and adds the tooling.

  • wayfinder-maps status <dir> — what is resolved, what is in flight, and what is ready to claim (the frontier: open, unclaimed, every blocker resolved).
  • wayfinder-maps lint <dir> — does the map still tell the truth?
  • wayfinder-maps serve / app — the map as a star-map, in a browser or a native window (design notes).
$ wayfinder-maps status ../expensif/.plan/daily-timeline
Daily Timeline — continuous days, empty days included
6 resolved · 0 claimed · 4 open · 0 out of scope

Frontier — ready to claim, first by number wins:
  05  Should HandleDaily's two branches converge      grilling
  06  Contain the day-card chrome drift               grilling
  07  The infinite-scroll island's contract           grilling
  10  Test strategy for the date-indexed timeline     grilling

Undermined — resolved on a premise that later changed:
  02  Window size and how older days load             broken by 08

Fog: 5 patches, 1 anchored to a ticket

Install

Skills, as a Claude Code plugin

/plugin marketplace add rengwu/wayfinder-maps
/plugin install wayfinder-maps@wayfinder-maps

/wayfinder-maps then works in every project; the map lands in that project's .plan/. The plugin bundles wayfinder-maps plus the four skills it invokes by name — grill-me, research, prototype, domain-modeling. All five were originally authored by Matt Pocock in mattpocock/skills; the copies under skills/ are MIT-licensed adaptations.

Skills, drop-in for any harness

Copy skills/ into your project and tell your agent to read skills/wayfinder-maps/SKILL.md and follow it; the skills it names live alongside. You may include any of the extras under skills/.optional/ too — skills/README.md describes the full set.

The binary

  • Download from Releases — no Go toolchain needed, native window and folder dialog on all three platforms. macOS uses WKWebView (part of the OS); Windows uses WebView2 (ships with Windows 10/11); the Linux archive bundles a small helper that binds the system webkit — libwebkit2gtk-4.1, preinstalled on most desktops, one sudo apt install libwebkit2gtk-4.1-0 otherwise. status and lint need nothing installed at all, even headless. The macOS and Linux archives also carry two double-clickable launchers — wayfinder-app opens the native window, wayfinder-web starts the server and opens your browser — no terminal needed (.command on macOS).
  • go install github.com/rengwu/wayfinder-maps/cmd/wayfinder-maps@latest

Skill and binary are independent — each just reads the on-disk contract.

Why

The map is shared memory: sessions orient to it before choosing work, and when it drifts it lies to them silently. The skill makes most drift unrepresentable and verifies each session's delta, no tool assumed. What a delta check cannot see is an edit made outside the protocol, or two parallel sessions taking the same ticket number. lint is fsck for those: it runs after the fact and re-establishes the base the next session rests on.

The format contract

The contract is the skill's local-markdown adapter, TRACKER-MARKDOWN.md — the skill itself is tracker-agnostic method. This tool implements that adapter; any other reader should target the same file. In summary: structure the facts, leave the prose alone. A ticket's type and edges have exactly one correct value a machine can check. Its Question and Answer, the map's Destination and Notes, and each one-line gist in Decisions-so-far are lossy prose, never parsed.

A ticket is .plan/<effort>/tickets/NN-<slug>.md:

---
type: research | prototype | grilling | task
blocked_by: [02, 03] # [] when none
claimed_by: <session id> # while a session holds it
claimed_at: 2026-07-10T09:00:00Z # RFC 3339, alongside claimed_by
undermined_by: [08] # optional — resolved on a premise 08 later broke
assets: [../assets/x.html.approved] # optional
---

# <Ticket title>

## Question

…

## Answer # its presence _is_ the resolution

…

Status is derived

There is no status: field — every value it could hold is already written in the file, and a second copy is free to go stale:

Derived statusWhen
resolvedthe body has an ## Answer with prose under it
out_of_scopethe body has a ## Ruled out with prose under it
claimedneither, and claimed_by is set
opennone of the above

Closure is read first, so a claimed_by left on a closed ticket is inert litter. And it is the prose, not the heading, that closes a ticket: a session that types ## Answer and dies has resolved nothing — the ticket stays claimed, and its claim goes stale on schedule. lint errors on the empty heading outright.

undermined_by marks a decision whose premise a later ticket broke; without it a renderer paints the node green and launders a live problem into a checkmark. claimed_at separates a dead session from live work. And out_of_scope stays distinct from resolved — a scope boundary is not a step on the route, and a parser that lumps them in over-counts.

Fenced code blocks are not structure

Every scan — headings, titles, bullets, links — ignores what sits inside a ``` or ~~~ fence. A ticket that quotes the ticket format contains the line ## Answer, and must not thereby resolve itself. The rule cuts one way only: an answer whose entire body is a code fence is still an answer. Any other reader of these files owes the same rule.

One session at a time

The adapter forbids concurrent sessions: git merges a duplicate ticket number cleanly and neither session can see it. lint still reports duplicates, because a merge can happen anyway.

Fog patches in the map's ## Not yet specified are bullets with a bolded lead title and an optional anchor; the rest stays loose prose:

- **Today's row.** Whether today is visually distinguished. <clears-with: 04>

Checks

Errors: dangling or self blocked_by; blocked_by cycles; duplicate ticket numbers; a ticket carrying both ## Answer and ## Ruled out; a closing heading with nothing under it; a resolved ticket missing from Decisions-so-far; an out-of-scope ticket listed as a decision, or missing from Out-of-scope; Decisions-so-far pointing at an unresolved ticket; a fog patch duplicating a live ticket's title, or anchored to a resolved ticket; a missing Destination; an unknown type; a leftover status: field that disagrees with the body.

Warnings: loose (pre-frontmatter) headers; a status: field that merely duplicates the body; a claim with no owner or older than 72h; a claim left on a closed ticket; a ticket blocked by something out of scope, which can never unblock.

Checks that need judgment — does a gist say what its answer says, does a fog title duplicate a live ticket in substance — stay with the skill; no parser can make them.

Exit 0 clean, 1 errors found, 2 no map at that path.

Reading legacy maps

Two older shapes still parse, each lints as a warning, and neither is written any more:

  • the loose header (Type: / Status: / Blocked by: after the H1), from before frontmatter;
  • a stored status: field, from before status was derived. Where it agrees with the body, lint says to delete it; where it disagrees, that is an error and the body wins.

Build from source

go build ./cmd/wayfinder-maps      # pure Go: status, lint, serve
go test ./...

The native window needs cgo on macOS (WKWebView); Windows is pure Go (WebView2). On Linux the webkit linkage lives in a separate helper, ./cmd/wayfinder-maps-webview, built with libwebkit2gtk-4.1-dev and kept next to the main binary. Releases are cut by tagging v* — CI builds the helper on Linux runners and GoReleaser packs everything.

常见问题

What is wayfinder-maps?

wayfinder-maps is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by rengwu. Planetary visualization tool for /wayfinder maps. It has 93 GitHub stars.

Is wayfinder-maps safe to use?

Yes. wayfinder-maps passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.

How do I install wayfinder-maps?

Clone the repository with "git clone https://github.com/rengwu/wayfinder-maps" and add it to your Claude Code skills directory (see the Installation section above).

What programming language is wayfinder-maps written in?

wayfinder-maps is primarily written in Go. It is open-source under rengwu on GitHub, so you can review or fork the full source.

Are there alternatives to wayfinder-maps?

Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh wayfinder-maps against similar tools.

评论 (0)

暂无评论,成为第一个分享想法的人!

ECC

by affaan-m

10

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

242,21936,702JavaScript
AI 智能体ai-agentsanthropicclaude-code
查看详情
15

An agentic skills framework & software development methodology that works.

234,96620,863Shell
AI 智能体ai-agentsbrainstorming
查看详情

hermes-agent

by NousResearch

10

The agent that grows with you

234,43747,175Python
AI 智能体ai-agentsagent-orchestration
查看详情

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

185,94028,768JavaScript
AI 智能体ai-agentsanthropicclaude-code
查看详情

cc-switch

by farion1231

3

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io

128,8688,826Rust
AI 智能体claude-codeai-tools
查看详情

claude-code

by anthropics

Claude Code is an agentic coding tool that lives in your terminal, understands your codebase, and helps you code faster by executing routine tasks, explaining complex code, and handling git workflows - all through natural language commands.

120,03119,897Shell
AI 智能体
查看详情

开发者还喜欢

基于喜欢此 Skill 的开发者投票和收藏

ECC

by affaan-m

10

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

242,21936,702JavaScript
AI 智能体ai-agentsanthropicclaude-code
查看详情
15

An agentic skills framework & software development methodology that works.

234,96620,863Shell
AI 智能体ai-agentsbrainstorming
查看详情

hermes-agent

by NousResearch

10

The agent that grows with you

234,43747,175Python
AI 智能体ai-agentsagent-orchestration
查看详情

n8n

by n8n-io

12

Fair-code workflow automation platform with native AI capabilities. Combine visual building with custom code, self-host or cloud, 400+ integrations.

201,88160,308TypeScript
MCP 服务器apisai-tools
查看详情

The agent harness performance optimization system. Skills, instincts, memory, security, and research-first development for Claude Code, Codex, Opencode, Cursor and beyond.

185,94028,768JavaScript
AI 智能体ai-agentsanthropicclaude-code
查看详情

cc-switch

by farion1231

3

A cross-platform desktop All-in-One assistant for Claude Code, Codex, OpenCode, OpenClaw, Grok Build & Hermes Agent. Only official website: ccswitch.io

128,8688,826Rust
AI 智能体claude-codeai-tools
查看详情