wayfinder
Skill173 starsChanged 17 days agoMITView on GitHub
Wayfind a foggy, multi-session effort into an accepted repository spec and separate implementation-ticket drafts. Use when the destination is known but important decisions, dependencies, or scope are not yet clear enough to implement safely.
194 other projects use a version of this.
Adopted by
Most recently taken up by
- therealtinhtute/mono-harnesstheir own version, first seen Nov 8, 2026
- domengabrovsek/agent-configtheir own version, first seen Nov 8, 2026
- domengabrovsek/claudetheir own version, first seen Nov 8, 2026
- The-Vibe-Company/quivrtheir own version, first seen Nov 8, 2026
- 5thlayer/skillworkstheir own version, first seen Nov 8, 2026
- zhikedan/mattpocock-skills-dshtheir own version, first seen Nov 8, 2026
Used with
Projects with it also use setup-ts-deep-modules, to-questionnaire and 10 more.
Other files people use alongside this one.
Files found in the same public projects as a version of this one, more often than they turn up anywhere else. Each project counts once, large collections count for less, and this file's own repository is left out.
liufeng1976Skillsetup-ts-deep-modules
Wire dependency-cruiser into a TypeScript repo so each package is a deep module — implementation hidden in subfolders, reachable only through its entry-point files. User-invoked.
Also in 42 projects that use this
AWSSkillto-questionnaire
Turn a decision you can't fully answer into a questionnaire for someone else to fill in.
Also in 87 projects that use this
AWSSkillwizard
Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.
Also in 124 projects that use this
RoentekSkillloop-me
Grill me about specs for the workflows I want to build, within this workspace.
Also in 50 projects that use this
LesaniSkillwriting-beats
Shape an article as a journey of beats, choose-your-own-adventure style. The user picks a starting beat from the raw material, you write only that beat, then offer options for where to pivot next, beat by beat, until the article reaches a natural end. Use when the user has raw material and wants to assemble it as a narrative rather than an argument.
Also in 54 projects that use this
ryotaroh180105Skillclaude-handoff
Hand the current conversation off to a fresh background agent that picks up the work immediately.
Also in 40 projects that use this
- +2 more from aws
Other versions194 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 194
- mattpocock/skills★ 309ksame name
- QuivrHQ/quivr★ 40ksame name
- The-Vibe-Company/quivr★ 40ksame name
- ArcReel/ArcReel★ 6.3ksame name
- vinvcn/mattpocock-skills-zh-CN★ 5.5ksame name
- withcoral/coral★ 4.9ksame name
- epicenter-md/epicenter★ 4.8ksame name
- EpicenterHQ/epicenter★ 4.8ksame name
- bestofjs/bestofjs★ 3.1ksame name
- CelestoAI/SmolVM★ 1.2ksame name
184 more not listed
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- autonomous-runSkill
- example-skillSkill
- grill-with-docsSkill
- handoffSkill
- harness-doctorSkill
- harness-helpSkill
- new-feedbackSkill
- new-researchSkill
- verifySkill
- writing-rulesSkill
The file
We split this skill into the rules it gives an agent
11 rules. 0 are common, 7 are distinctive, and 4 are very specific to this project.
0
Common
Most of the 179 edited versions we read keep these.
7
Distinctive
This author's own choices. Worth a look.
4
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 55
When unsure, take the heavier route.
- Very specific: line 59
If the superpowers plugin is enabled, use its skills only for a stage: this order wins over its session-start instruction.
- Very specific: line 66
Read the tracked item, existing plans/specs, and docs/14-project-tracker-guide.md when present.
- Very specific: line 70
Create docs/specs/<slug>.md from .harness/templates/wayfinder-spec.md (or templates/wayfinder-spec.md in the harness source checkout).
- Distinctive: line 73
Add every sharp question to the decision map, wire its blockers, and identify the frontier.
- Distinctive: line 74
Keep unshaped uncertainty in Fog.
- Distinctive: line 75
If tracker writes are unauthorized, put paste-ready ticket bodies in the spec and surface them to the operator.
- Distinctive: line 77
Stop after charting. The completed draft spec and ticket drafts are the evidence for this step.
- Distinctive: line 86
Add only a linked gist to the Decision index.
- Distinctive: line 89
Stop after one decision in attended mode.
- Distinctive: line 97
no unresolved in-scope decision or blocking fog remains.
“Common” is measured: kept by at least 6 in 10 of the 179 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Wayfinder
Turn uncertainty into a decision-complete spec. The map plans the route; it does not execute the destination.
Invariants
- Destination: the observable end state this effort is finding a route to.
- Decision map: named questions and their dependency edges. A decision produces an answer, not a slice of implementation.
- Frontier: the sharp, unresolved, unblocked decisions that can be taken next.
- Fog: in-scope uncertainty that cannot yet be phrased as a precise question. Fog graduates to a decision only when the question is sharp.
- Decision index: one context pointer and one-line rationale per resolved decision; the detail lives in its ticket or linked research artifact.
- Non-goal: a deliberate scope boundary. It never graduates from fog unless the destination is explicitly redrawn.
Attended mode (default): resolve one decision per session with the operator. Parallel research may inform that decision, but does not silently resolve another one.
Unattended mode (a batch without the operator): resolve several decisions in one session when an agent can answer them, and record the rationale of each one separately. A decision that needs the human's preference is never answered by an agent: put it under "Needs operator" with the decisions it blocks. It blocks only those; the rest of the map moves on.
Pipeline and route
Wayfinder covers part of the development pipeline. Each stage leaves an artifact:
- Goal: the goal and problem restated in the spec.
- Research:
docs/research/<slug>.md, premise reproduced; named inresearch:. - Shape: the draft spec in
docs/specs/<slug>.md. - Route:
route:with a one-line reason. - Spec review: a verdict file from a reviewer who did not draft the spec, named in
reviewer_receipt:, plus the "Review resolution" section. - Accept: the operator only.
- Implement:
autonomous-run, or an attended session. - Prove: verifier output and
check-specsoutput, both shown. - Land: commit or pull request with the evidence, by the operator's delivery rule.
Pick the route and record why:
direct: decision-complete, one session, clear "done" (a reproduced bug, a docs fix).grill: unclear shape, one session, the operator is present; usegrill-with-docs.wayfinder: two or more dependent open decisions, several sessions, several subsystems or owners, or a change to a data model, security boundary or failure mode.
When unsure, take the heavier route. direct still needs research, a spec review and acceptance.
A list of bugs is a numbered checklist of direct items, not one pipeline item. AgentSmith's
docs/28-development-pipeline.md gives the reasons.
If the superpowers plugin is enabled, use its skills only for a stage: this order wins over its
session-start instruction; save its spec as docs/specs/<slug>.md; after brainstorming go to
Route, not writing-plans; in an unattended batch do not wait for its human review, use the
unattended mode.
Chart
- Read the tracked item, existing plans/specs, and
docs/14-project-tracker-guide.mdwhen present. Determine the active tracker-write policy before any tracker action. - Name the destination and non-goals, and choose the route as above. Continue here only for
wayfinder. - Create
docs/specs/<slug>.mdfrom.harness/templates/wayfinder-spec.md(ortemplates/wayfinder-spec.mdin the harness source checkout); reproduce that shape only when neither is available. Initial status is alwaysdraft. Fillresearch:androute:. - Add every sharp question to the decision map, wire its blockers, and identify the frontier. Keep unshaped uncertainty in Fog. Draft a decision ticket for each mapped decision.
- If tracker writes are unauthorized, put paste-ready ticket bodies in the spec and surface them to the operator. A connected or named tracker grants no write permission.
- Stop after charting. The completed draft spec and ticket drafts are the evidence for this step.
Advance
- Load the spec at low resolution: destination, frontier, fog, non-goals, and decision index.
- Select one frontier decision (several in unattended mode). Research, prototype, or discuss only enough to answer its stated question. Human-preference decisions require the human's actual answer; in unattended mode they go under "Needs operator".
- Record the answer and rationale once, in the decision's ticket draft or authorized tracker item. Add only a linked gist to the Decision index.
- Recompute dependencies. Graduate newly sharp fog into decisions; move newly excluded work to Non-goals or Explicit deferrals with a reason.
- Stop after one decision in attended mode. Completion means each answer is recorded and the map reflects the new frontier without duplicate rationale.
Terminal-spec gate
A spec can be offered for acceptance only when:
- the destination and non-goals are concrete;
- no unresolved in-scope decision or blocking fog remains;
- each resolved decision has a rationale and context pointer;
- evidence/acceptance requirements cover the destination end to end;
- implementation-ticket drafts are separate, independently executable scopes; and
research:,route:andreviewer_receipt:are filled and both cited files are committed (a run's worktree has only what Git has), the reviewer did not draft the spec, "Review resolution" answers each finding, andpython3 .agentsmith/autonomous-run.py check-specs docs/specs/<slug>.mdexits 0.
Present the spec as draft and ask the operator to accept or revise it. Only an explicit operator
acceptance may change status to accepted; record that human in accepted_by and the acceptance
time in accepted_at. An agent's own evaluation cannot accept it. Acceptance closes the decision
item and unlocks planning/execution against the separate implementation tickets.
Tracker consent still governs closing or creating items: when writes are unauthorized, provide the
close comment and implementation-ticket bodies as paste-ready drafts.
Implementation boundary
An implementation ticket names one deliverable, its scope, dependencies, and proof. It never
inherits the decision ticket's identity. Wayfinder ends at the accepted spec and ticket drafts;
start implementation only from a separately tracked implementation ticket and the operator's
normal execution instruction. The usual successor is autonomous-run, which implements one such
ticket against the accepted spec and refuses a spec that fails check-specs.
Adapted from Matt Pocock's
wayfinder
(MIT). Agentsmith keeps its decision-map concepts while making the terminal artifact repository-
native and preserving tracker consent.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
Reviews can't be read right now.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

