One-prompt build · 2026-10-24
Skills Library
Paste this into a new Shogo project and watch it build.
Build a single-page **skills library** for an agent dev team: a planner, three developers and a QA
agent. It will be screen-recorded at phone width (390px) and at desktop width, so it must look
finished at both. Ship it running in the preview, then reply with the preview URL and how many
minutes it took.
Dark UI, near-black background, one column on phone and a two-column grid on desktop, mono labels,
big type for the headings.
**Hero**
- small mono label: `SKILLS · the rules our agents are graded against`
- H1: `Twelve skills. Three roles. One merge.`
- one line under it: `These are the rules our planner, developer and QA agents actually run, in the order a card hits them.`
**Role filter** — four chips above the cards: `All 12` · `Planner 4` · `Dev 4` · `QA 4`.
Clicking a chip filters the list. `All 12` is active on load.
**The twelve cards**, in this order, each with a role tag, a number, a title, one line on when to use
it, and the rule itself in a copy block:
Planner
1. `PLAN-01 · Read the post, then find the files` — before the plan, not after. Rule: `Find the real files first. A plan that names imaginary files is a guess.`
2. `PLAN-02 · The plan names what done looks like` — every card, no exceptions. Rule: `problem · how to reproduce · approach · acceptance criteria · test plan.`
3. `PLAN-03 · Ask, don't guess` — when the outcome could go two ways. Rule: `Up to three questions, recommended answer first, and the card waits in Needs Info.`
4. `PLAN-04 · One idea, one card` — never split a feature. Rule: `One plan, one developer, one branch. Three cards became three PRs that did not fit together.`
Dev
5. `DEV-01 · Build only what the card says` — scope is the card. Rule: `The card is the spec. Anything not on it is a different card.`
6. `DEV-02 · Reproduce before you fix` — every bug card. Rule: `A failing test that reproduces the bug, then the fix, then the test passing.`
7. `DEV-03 · One active card per developer` — always. Rule: `A card in Changes Requested still counts as active. Two cards fight over one checkout.`
8. `DEV-04 · Never arm auto-merge early` — QA opens the PR. Rule: `Auto-merge armed before QA is how four open blockers reached main.`
QA
9. `QA-01 · Read the diff against the checklist` — every card. Rule: `No dead code, no debug logs, no comments narrating the code.`
10. `QA-02 · Write the tests that are missing` — when the developer skipped them. Rule: `Integration and end-to-end where the diff needs it, written by QA, run by QA.`
11. `QA-03 · A test that passes without the fix proves nothing` — always. Rule: `Run the new test on main first. If it passes there, send the card back.`
12. `QA-04 · Number your findings` — every round. Rule: `Round 1: Changes Requested with numbered findings. Round 2: the verdict, then the PR.`
**Closing block** — the biggest type on the page:
`twelve skills · three roles · one human merge`
Under it, small: `the gate is a different agent from the one that wrote the code.`
**Footer chip**: `built in Shogo from one prompt`
Rules: no charts, no fake metrics, no invented numbers, no placeholder copy blocks. The copy blocks
are real text a reader can select. One screen; keep the hero short enough that the first two cards
show at phone width.Want it as a guided course with a coach?Agent Team Playbook