agentleFS
EastkapEastkap/autocomp/.claude/skills/next-move/SKILL.md

next-move

Skill4 starsChanged 5 months agoMITView on GitHub

Decide the single highest-leverage next thing to build or do — to close the gap to our competitors (Polsia, Pancake, NanoCorp) and reach the next revenue milestone (first signup, first sale, launch). Use when the owner asks "what's next", "what should we build next", "what gets us closer to <rival>", "how do we get our first sale", or when a tick/session needs to re-pick priorities instead of just draining the backlog top-down. Produces ONE pick with a verifiable success criterion + the honest binding constraint, not a list.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 6 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Look at the KPIs and ask: what is the ONE thing whose absence is blocking the milestone right now?

    line 35
  2. Distinctive:

    If signups are ~0 but the page works and converts fine on the humans who arrive → the constraint is distribution, not the product.

    line 38
  3. Distinctive:

    If traffic is fine but nobody trusts it → the constraint is credibility (proof, /live feed, verified revenue), build that.

    line 40
  4. Distinctive:

    If the launch is what unlocks eyeballs → the constraint is launch-readiness, build the artifact the launch needs, then the launch itself is …

    line 42
  5. Very specific:

    Run tools/scoreboard.sh (tokens/$ burned vs income per venture).

    line 57
  6. Distinctive:

    Buildable-now breaks ties, never overrides §1: If the top constraint-mover is gated (a send), the pick is: *do the full buildable prep now, …

    line 72
  7. Distinctive:

    Table-stakes before differentiators before polish: If we lack something the whole category has (e.g.

    line 77

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

next-move — pick the one highest-leverage thing to build next

The recurring co-founder question: given where our rivals are and where our money is, what is the single most valuable thing to do next? This skill turns that instinct into a repeatable decision. It does not drain the backlog top-down — it re-derives the pick from reality each time, because the binding constraint moves.

Core discipline (why this exists): the honest answer is often "the highest-leverage move is gated (a send / a spend), and no amount of building substitutes for it." A loop that can only build will happily build forever to avoid the real blocker. This skill's job is to name that blocker out loud, then pick the best move given it — buildable now if one clears the bar, boarded for the human if the real lever is theirs to pull.

1. Ground in reality (read, don't recall)

Load the current picture — never answer from memory:

  • private/memory/competitive-landscape-*.md — where rivals are, table stakes, our open gap.
  • private/charter.md — the venture, the pricing, and the Definition of success (the goal we're steering toward; that's the destination, not "beat rival X" in the abstract).
  • private/state/kpis.md — what's actually measured (signups, visits, revenue). The truth.
  • private/state/backlog.md + private/state/pnl.md — what's queued, what's spent, runway.
  • tools/tasks.sh list — what's already on the human's plate (don't re-board it).

2. Name the destination concretely

"Closer to Polsia/Pancake/NanoCorp" is not a goal — it's a direction. Convert it to the next measurable milestone on the path: usually the charter's Definition of success, or the nearest sub-step (first human signup → first pre-order → launch that produces N signups). Write it as one line with a number. Everything below is judged against this, not against vibes.

3. Find the binding constraint (the honest part)

Look at the KPIs and ask: what is the ONE thing whose absence is blocking the milestone right now? Distribution? Trust/credibility? A missing table-stakes feature? Product itself? Be ruthless — there is usually exactly one, and it's usually not the thing that's most fun to build.

  • If signups are ~0 but the page works and converts fine on the humans who arrive → the constraint is distribution, not the product. More features won't move it.
  • If traffic is fine but nobody trusts it → the constraint is credibility (proof, /live feed, verified revenue), build that.
  • If the launch is what unlocks eyeballs → the constraint is launch-readiness, build the artifact the launch needs, then the launch itself is the (gated) move.

4. Enumerate candidate moves — tag each

List 3–6 real candidates (pull from the backlog + the competitive gap + the binding constraint). For each, tag four things in one line:

  • Unlocks — what milestone-relevant thing it produces (be specific, tie to §2).
  • Buildable | Gated — can the loop ship it with the keys it holds (Cloudflare/Supabase/repo/ deploy), or does it need a human gate (spend / outbound send / their-identity credential / judgment)? (CLAUDE.md hard rules + #12.)
  • Effort — rough (an hour / a tick / multi-tick).
  • Leverage — does it move the binding constraint (§3), or a secondary one? Table-stakes-we- lack outrank nice-to-haves; a thing the next move depends on outranks a thing nothing waits on.

4.5 The Rule-17 gate (distribution-first — check BEFORE you let "build" win)

Run tools/scoreboard.sh (tokens/$ burned vs income per venture). Then apply CLAUDE.md #17: a new venture, or a new feature/page/tool, is DISQUALIFIED as the pick while any active venture still has loop-doable GTM undone or no stats flowing. Building is the tar pit; selling what exists is the job — the best ad is our own company actually converting. If the scoreboard shows a SELL verdict (burn, $0 in, no live GTM), the pick MUST come from the distribution set (ship+submit SEO, GEO, directories, draft+board outreach), never the construction set. Owner-gated distribution (sends/spend) counts as "done on our side" once boarded — that unblocks building, a still-unworked GTM lever does not. Only after every active venture clears this do build-candidates re-enter §5.

5. Pick ONE — apply the tie-breakers in order

  1. Binding-constraint first. A move that clears §3 beats any move that doesn't, even a smaller one. Don't build a dashboard when distribution is the wall. (And per §4.5, a build candidate can't even be the pick while GTM on a live venture is unworked.)
  2. Hundred-Dollar Test (CLAUDE.md #6). Of the constraint-movers, which is the highest- leverage hour/$100 right now? Cheapest path to the milestone wins.
  3. Buildable-now breaks ties, never overrides §1. If the top constraint-mover is gated (a send), the pick is: do the full buildable prep now, and board the irreducible gated remainder for the human (CLAUDE.md #12). Do NOT substitute a lower-leverage buildable thing and call it the next move — that's building to dodge the blocker. Name the gated lever as the real next move; the buildable prep is you doing your half.
  4. Table-stakes before differentiators before polish. If we lack something the whole category has (e.g. a public /live feed), that gap costs credibility on every visit — it usually outranks a novel feature no one's asking for yet.

6. Output — the pick, not a menu

Return exactly:

  • The binding constraint (§3) in one honest sentence.
  • The pick — one move, with its verifiable success criterion (Principle 4: "page returns 200 and shows X", "row lands in table Y", not "build Z").
  • Why it wins — 2–3 bullets tying it to the constraint + the tie-breakers.
  • The gated lever, named — if the true highest-leverage move needs the human (a send/spend), say so plainly and state what buildable half you'll do now vs. what goes on the board.
  • Runner-up — one line, so the owner can override.
  • Then act (CLAUDE.md #12/#16): do the buildable half now toward its success criterion, independently verify it (tools/verify-goal.md), and board the irreducible remainder via tools/tasks.sh add. Anything you touched lands on the kanban with the right status.

Anti-patterns (this skill exists to prevent these)

  • Backlog autopilot — doing the top backlog item because it's on top, when the constraint moved underneath it.
  • Building to avoid the blocker — shipping a fifth SEO page when the wall is that nobody sees the site (distribution is gated, so we build instead of asking). Name the gate.
  • Menu instead of a decision — listing six options and letting the owner pick is abdication. Pick one, show the runner-up, invite override.
  • Vanity table-stakes over the real constraint — building the /live feed because rivals have it, when we have zero traffic to be live about. Table-stakes matter, but not ahead of §3.
  • A "build" as the success criterion — "build the dashboard" is not verifiable; "hq.limed.tech loads authed and every number traces to a real source" is.

Cost note

This is a ~few-minute reasoning pass over files already on disk. Run it at the start of a tick when the backlog and reality have drifted, or any time the owner asks "what's next" — it's cheaper than a tick spent building the wrong thing.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

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.