agentsclimarketplace

Wayfinding

Skill lawzava/megapowers/plugins/mega-orchestration/skills/wayfinding

Skills, plugins, and hooks that change how coding agents work — Claude Code, Codex, OpenCode, Antigravity. Measured, not vibed.

Install
npx -y skills add lawzava/megapowers --skill wayfinding

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when long-horizon work has unknown ownership, unresolved decisions, or unclear sequencing that prevents an honest specification or implementation plan.

The file declares its own license as MIT. That is the author’s claim about this one file, and it is not the same thing as the license GitHub reports for the repository, which is listed with the other numbers below.

SKILL.md

2.5 KB, as published. Nobody here has run it

Wayfinding

Reduce uncertainty until the work is ready for design or planning. Wayfinding is discovery and decision resolution, not delivery.

Invocation boundary

On Codex, agents/openai.yaml disables implicit invocation: invoke $wayfinding explicitly. Other harnesses consume the portable frontmatter and may still invoke this skill implicitly. The explicit-only policy is Codex-only.

Local contract

Use .megapowers/wayfinding/<id>/ by default:

  • .megapowers/wayfinding/<id>/map.md: outcome, known regions, fog, source trust, owners, named decisions, evidence, dependencies, and the current frontier.
  • decisions/<decision-id>.md: one question, its owner, evidence and source trust, options, status, and consequences for the map.

The default .megapowers/ path is ignored personal workspace state. If the user wants a shared committed map, agree on a project-owned path first and use the same contract there. Never force ignored artifacts into git. Mirroring the map into an issue tracker is optional; the local artifacts remain complete without a tracker.

Loop

  1. Read the strongest available repository and external sources. Record what each source establishes and how much it should be trusted.
  2. Create or refresh map.md. Separate known regions from fog, name owners and unresolved decisions, record dependencies, and mark one current frontier.
  3. Resolve one decision at a time. Write or update its decision file with the evidence, options, status, and map consequences.
  4. Update the map after each decision: propagate consequences, revise fog and dependencies, then update the current frontier.

Never implement, execute a plan, start an autonomous run, or automatically commit while wayfinding.

Stop and hand off

  • Spec-ready: the outcome, constraints, behavior, and success criteria can be designed without inventing missing facts. Hand off to megapowers:brainstorming.
  • Plan-ready: plan-ready is valid only when an approved design already exists and the remaining uncertainty no longer prevents honest decomposition. Hand off to megapowers:writing-plans.
  • Blocked: name the missing external evidence, its owner, and what would unblock the current frontier. Leave the map ready to resume.

These are terminal states. Do not hand directly to implementation or an autonomous run.

Keep looking

Skills are one crate of 328,083. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.