Skip to content
Open
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
42 changes: 21 additions & 21 deletions diagrams/skill-map.json
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,7 @@
"label": "/super-board",
"kind": "public",
"what": "The **autonomous GitHub-Project pipeline**: five verbs (onboard, lint, status, run, stop) take one board shape (Backlog · Ready · Building · QA · Review · Blocked · Done) ==from empty to drained==, with labels qa · bug · feature routing each card.",
"when": "Eric types **/super-board <verb>**, or says \"==drain my GitHub project==\" or \"stop super-board\"; with no verb it asks which one.",
"when": "The owner types **/super-board <verb>**, or says \"==drain my GitHub project==\" or \"stop super-board\"; with no verb it asks which one.",
"how": "The session is an **orchestrator** only: it routes the verb to its reference file, dispatches lane agents, and ==never does product work itself==. Board state is the only state, so `run` is also resume. Every commit, ticket, PR body and comment follows references/writing-standard.md.",
"source": [
"_skills/vendor/super-board/skills/super-board/SKILL.md",
Expand Down Expand Up @@ -62,8 +62,8 @@
"label": "/ui-refine-loop",
"kind": "public",
"what": "Unattended **Impeccable** check-then-fix loop on one page or component that ==ends in a draft PR== with a Before | After table. Standalone: no board lane ever starts it.",
"when": "Eric runs **/ui-refine-loop** <route | component | description> [\"what's wrong\"] [--rounds N], or says \"refine loop\" or \"==polish this page with impeccable==\".",
"how": "Reads the target's code, grills Eric (at most 10 questions) into a **taste file**, then runs the saved ui-refine-loop workflow: each round fresh checkers rank typed problems and a fresh fixer chains Impeccable verbs, polish last. Default 5 rounds; ==a finish reviewer grades every fix==. Commits are `💄 [ui] <slug>: …` and the PR body uses super-board-pr-body.sh's marker blocks.",
"when": "The owner runs **/ui-refine-loop** <route | component | description> [\"what's wrong\"] [--rounds N], or says \"refine loop\" or \"==polish this page with impeccable==\".",
"how": "Reads the target's code, grills the owner (at most 10 questions) into a **taste file**, then runs the saved ui-refine-loop workflow: each round fresh checkers rank typed problems and a fresh fixer chains Impeccable verbs, polish last. Default 5 rounds; ==a finish reviewer grades every fix==. Commits are `💄 [ui] <slug>: …` and the PR body uses super-board-pr-body.sh's marker blocks.",
"source": [
"_skills/vendor/super-board/skills/ui-refine-loop/SKILL.md",
"_skills/vendor/super-board/workflows/ui-refine-loop.js"
Expand All @@ -81,7 +81,7 @@
"label": "/visual",
"kind": "public",
"what": "Turns the current work into ==one self-contained HTML page==. It can be a **branch recap**, a **plan**, or a map of a folder or architecture, with diagrams, a file map and annotated code.",
"when": "Eric runs **/visual**, /visual recap [base], /visual plan [file] or /visual <path>; ==with no argument visual.py detect picks the mode==.",
"when": "The owner runs **/visual**, /visual recap [base], /visual plan [file] or /visual <path>; ==with no argument visual.py detect picks the mode==.",
"how": "Writes a **data.json** from git facts or the plan, then visual.py render injects it into an inline template with ==no build step==. It screenshots the page headless to check it, and offers to publish only on a yes.",
"source": [
"_skills/vendor/super-board/skills/visual/SKILL.md"
Expand Down Expand Up @@ -134,7 +134,7 @@
"label": "/super-board onboard",
"kind": "public",
"what": "An 8-step setup wizard (Checks · GitHub · Board · Branch · AGENTS.md · Policies · Bug sources · Review) that installs or repairs what the board needs and writes the config once.",
"when": "Eric runs /super-board onboard (or /super-board:super-board onboard after a plugin install), on a new project, after an upgrade from 2.x, or to check and repair an existing setup.",
"when": "The owner runs /super-board onboard (or /super-board:super-board onboard after a plugin install), on a new project, after an upgrade from 2.x, or to check and repair an existing setup.",
"how": "One AskUserQuestion per step with the recommended option first; must-haves are fixed without asking by super-board-setup.py. Answers save to onboard-answers.json as they come, so a halt resumes, and repo files change only at step 8 after \"Write everything\". It ends by naming /super-board lint.",
"source": [
"_skills/vendor/super-board/skills/super-board/references/onboard.md",
Expand All @@ -148,8 +148,8 @@
"label": "/super-board lint",
"kind": "public",
"what": "Pre-flight readiness check that walks Ready, Building, QA and Review, flags issues that fail any of 18 criteria, proposes a qa/bug/feature label, and dispatches a helper skill per flagged issue.",
"when": "Eric runs /super-board lint before a run, typically after /super-collect filed new Backlog cards.",
"how": "Criteria 13–18 cover what the loop needs: a parseable `## Blocked by` line, a fake with no follow-up card, too big (over ~400 changed lines or many areas), the ticket sections, a checklist AC, and lettered Context steps. Phase 4 routes each flagged pattern to one skill sub-agent while Eric keeps the pace.",
"when": "The owner runs /super-board lint before a run, typically after /super-collect filed new Backlog cards.",
"how": "Criteria 13–18 cover what the loop needs: a parseable `## Blocked by` line, a fake with no follow-up card, too big (over ~400 changed lines or many areas), the ticket sections, a checklist AC, and lettered Context steps. Phase 4 routes each flagged pattern to one skill sub-agent while the owner keeps the pace.",
"source": [
"_skills/vendor/super-board/skills/super-board/references/lint.md"
],
Expand All @@ -161,7 +161,7 @@
"label": "/super-board status",
"kind": "public",
"what": "Read-only snapshot of the active config, every column in board order (Backlog included), and in-flight workers.",
"when": "Eric runs /super-board status to see where a run stands.",
"when": "The owner runs /super-board status to see where a run stands.",
"how": "super-board-status.py reads the board and local in-flight state and prints the columns in the board's own order; it changes nothing.",
"source": [
"_skills/vendor/super-board/skills/super-board/references/status.md",
Expand All @@ -174,7 +174,7 @@
"label": "/super-board run",
"kind": "public",
"what": "The autonomous loop: repeated waves of the super-board-wave workflow from this session until the board drains or a halt gate fires. Also the resume command.",
"when": "Eric runs /super-board run [--low|--high] [<slug>]; the default worker_backend \"workflow\" stays in-session, \"claude-p\" opts into the legacy headless runner.",
"when": "The owner runs /super-board run [--low|--high] [<slug>]; the default worker_backend \"workflow\" stays in-session, \"claude-p\" opts into the legacy headless runner.",
"how": "Each wave: rate and usage guard, plan with super-board-wave-plan.sh (which refuses an old `variant: qa-only` config with exit 65), act on sweep/flag/stranded/resume, claim cards by assignee, launch the workflow, reconcile, report. Halts after 3 zero-progress waves, a high block rate, or a stop.",
"source": [
"_skills/vendor/super-board/skills/super-board/references/run-workflow.md",
Expand All @@ -192,7 +192,7 @@
"label": "/super-board stop",
"kind": "public",
"what": "Graceful shutdown: stops the wave, posts \"stopped mid-flight\" comments, releases assignee claims, and removes the wave lock so the next run resumes.",
"when": "Eric runs /super-board stop or says \"stop the run\" / \"pause the loop\".",
"when": "The owner runs /super-board stop or says \"stop the run\" / \"pause the loop\".",
"how": "Stops the workflow (or kills the legacy dispatcher via super-board-stop.sh), comments on every in-flight issue and PR, and strips bot_identity assignees. Resume is just `run` again.",
"source": [
"_skills/vendor/super-board/skills/super-board/references/stop.md",
Expand Down Expand Up @@ -236,7 +236,7 @@
"label": "super-review",
"kind": "lane",
"what": "**Reviewer lane** and standalone PR review: forms its own hypotheses first, remembers prior findings, classes every finding, and ==merges only through the merge gate==.",
"when": "==Started by the board== for cards in **Review** (always on the session model). Also runnable standalone when Eric says \"review this branch\" or \"is this merge-ready\".",
"when": "==Started by the board== for cards in **Review** (always on the session model). Also runnable standalone when the owner says \"review this branch\" or \"is this merge-ready\".",
"how": "Thread scan, prior-report round 1, own pass with code-review, codebase-design and ponytail-review, a rerun of the Tester's exact test command, then an **adversarial truth-check** on non-trivial diffs. One report comment is edited in place every round. ==A clean review merges through the gate== pinned to the reviewed head SHA; exit 7 or 8 parks the card in Blocked with 🙋. Visual problems become a UI ticket, never a ui-refine-loop run. The truth-check sub-agents get a fixed verbatim brief, and Over-engineering is never a Blocker from any of them.",
"source": [
"_skills/vendor/super-board/skills/super-review/SKILL.md"
Expand Down Expand Up @@ -1473,7 +1473,7 @@
"label": "AGENTS.md + CLAUDE.md",
"kind": "board",
"what": "The project's instruction files: AGENTS.md as the single source of truth with a managed super-board block, and CLAUDE.md as `@AGENTS.md` plus Claude-only rules.",
"when": "Written at onboard step 8 when Eric chose to move his rules (or only add the section); the block is rewritten on re-install.",
"when": "Written at onboard step 8 when the owner chose to move their rules (or only add the section); the block is rewritten on re-install.",
"how": "The block sits between `<!-- super-board:begin -->` and `<!-- super-board:end -->`, rendered from agents-md-block.md with a Writing table that links the writing standard; nothing outside the markers is touched.",
"source": [
"_skills/vendor/super-board/skills/super-board/references/agents-md-block.md",
Expand Down Expand Up @@ -1519,7 +1519,7 @@
"label": "get.sh",
"kind": "script",
"what": "One-line installer: `curl -fsSL …/get.sh | bash` fetches the latest super-board release and runs install.sh.",
"when": "Eric (or a user) runs it in a project folder to install or upgrade; options --target, --ref, --no-helper-skills, and --no-hooks / --protect-main passed through.",
"when": "The owner runs it in a project folder to install or upgrade; options --target, --ref, --no-helper-skills, and --no-hooks / --protect-main passed through.",
"how": "Downloads the release tarball, runs install.sh, adds Matt Pocock's skills via `npx skills@latest add mattpocock/skills` unless told not to, and prints the next step once.",
"source": [
"_skills/vendor/super-board/get.sh"
Expand Down Expand Up @@ -1565,7 +1565,7 @@
"label": "/git-sync",
"kind": "public",
"what": "Commits the working tree as **one commit per logical change**, merge-pulls, pushes the current branch ==without force==, and reports what it did.",
"when": "Eric runs **/git-sync** or says \"sync\", \"commit and push\" or \"pull and push\", usually ==at the end of a piece of work==.",
"when": "The owner runs **/git-sync** or says \"sync\", \"commit and push\" or \"pull and push\", usually ==at the end of a piece of work==.",
"how": "Stages each group by path (never `git add -A`), **refuses secrets**, writes the host's commit format or the writing standard, gates `supabase/` changes on a migration file and advisors, then `git pull --no-rebase` and `git push`. ==The default or a protected branch needs a yes==.",
"source": [
"_skills/vendor/super-board/skills/git-sync/SKILL.md"
Expand Down Expand Up @@ -1724,35 +1724,35 @@
"from": "super-board",
"to": "sb-onboard",
"label": "routes onboard",
"when": "Eric runs /super-board onboard"
"when": "The owner runs /super-board onboard"
},
{
"id": "sb-v-lint",
"from": "super-board",
"to": "sb-lint",
"label": "routes lint",
"when": "Eric runs /super-board lint"
"when": "The owner runs /super-board lint"
},
{
"id": "sb-v-status",
"from": "super-board",
"to": "sb-status",
"label": "routes status",
"when": "Eric runs /super-board status"
"when": "The owner runs /super-board status"
},
{
"id": "sb-v-run",
"from": "super-board",
"to": "sb-run",
"label": "routes run",
"when": "Eric runs /super-board run, or asks to resume"
"when": "The owner runs /super-board run, or asks to resume"
},
{
"id": "sb-v-stop",
"from": "super-board",
"to": "sb-stop",
"label": "routes stop",
"when": "Eric runs /super-board stop or says pause"
"when": "The owner runs /super-board stop or says pause"
},
{
"id": "sb-onb-config",
Expand Down Expand Up @@ -1793,7 +1793,7 @@
"id": "lint-grill",
"from": "sb-lint",
"to": "grilling",
"label": "grills Eric",
"label": "grills the owner",
"when": "an issue needs a rethink (criteria 10, 11) or has several readings (12)"
},
{
Expand Down Expand Up @@ -2150,7 +2150,7 @@
"id": "u-grill",
"from": "ui-refine-loop",
"to": "grilling",
"label": "grills Eric",
"label": "grills the owner",
"when": "after reading the target's code; at most 10 questions in waves"
},
{
Expand Down
4 changes: 2 additions & 2 deletions diagrams/writing-standard/reference.html
Original file line number Diff line number Diff line change
Expand Up @@ -108,7 +108,7 @@ <h1>Writing standard</h1>
Co-Authored-By: Claude …</pre></div></div></section><section class="ref" id="status"><h2>PR status card</h2><p class="purpose">The first line of the PR. It says where the PR stands.</p><div class="card"><div class="chead">GitHub alert</div><div class="render">
<div class="alert"><div class="t">ⓘ Note</div>⏳ In review · round 1 · ✅ 2/3 AC · Closes #790</div>
<div class="alert tip"><div class="t">💡 Tip</div>✅ Merged · <code>9c41e07</code></div>
<div class="alert warn"><div class="t">⚠ Warning</div>🛑 Blocked · needs Eric</div>
<div class="alert warn"><div class="t">⚠ Warning</div>🛑 Blocked · needs owner</div>
<pre>&gt; [!NOTE]
&gt; ⏳ In review · round 1 · ✅ 2/3 AC · Closes #790</pre></div></div></section><section class="ref" id="problem"><h2>PR · Problem</h2><p class="purpose">Where it happens, who sees it, how to get there (one step per line), and what happened.</p><div class="card"><div class="chead">Lettered under Where</div><div class="render">
<h3>Problem</h3>
Expand All @@ -123,7 +123,7 @@ <h3>Solution</h3><ul><li>413 shows "File too large (max 10 MB)"</li><li>Client c
<h3>Acceptance criteria</h3><ul>
<li class="task"><input type="checkbox" checked disabled>12 MB file shows the size message in 1 s<span class="why">↳ <code>upload-size.spec.ts:14</code></span></li>
<li class="task"><input type="checkbox" checked disabled>No request over 10 MB<span class="why">↳ spec asserts 0 calls · HAR</span></li>
<li class="task"><input type="checkbox" disabled>No new Sentry events after deploy<span class="why">↳ not verified: needs prod deploy · Eric</span></li></ul></div></div></section><section class="ref" id="history"><h2>Iteration history</h2><p class="purpose">The lane timeline. GitHub markdown can't run JavaScript, so the time is fixed when the lane writes it.</p><div class="card"><div class="chead">Local time from config</div><div class="render">
<li class="task"><input type="checkbox" disabled>No new Sentry events after deploy<span class="why">↳ not verified: needs prod deploy · owner</span></li></ul></div></div></section><section class="ref" id="history"><h2>Iteration history</h2><p class="purpose">The lane timeline. GitHub markdown can't run JavaScript, so the time is fixed when the lane writes it.</p><div class="card"><div class="chead">Local time from config</div><div class="render">
<table><tr><th>Lane</th><th>Done</th><th>Time</th><th>Details</th></tr>
<tr><td>🔨 builder</td><td>✅</td><td>Oct 2, 10:02 EDT</td><td>draft</td></tr>
<tr><td>🔍 qa</td><td>❌ v1</td><td>Oct 2, 10:40 EDT</td><td>mobile hidden</td></tr>
Expand Down
4 changes: 2 additions & 2 deletions skills/super-board/references/block-template.md
Original file line number Diff line number Diff line change
Expand Up @@ -39,7 +39,7 @@ Why I (bot) cannot decide:
To unblock: <concrete action the human can take, in their own checklist form>
[ ] <step 1>
[ ] <step 2>
Owner: <who acts next — "Eric", "repo admin", "design" — never "someone">
Owner: <who acts next — "owner", "repo admin", "design" — never "someone">
Move back: drag this card to Ready after the steps above are done
blocked-by: <comma-separated issue numbers, or "-" if nothing on this board clears it>
```
Expand Down Expand Up @@ -116,7 +116,7 @@ After a trusted human confirms this version, the next wave returns it to Review
PR #<P> <one line — e.g. "adds prisma/migrations/0042_add_plan">. <env> isn't in the databases the robot may migrate.
Tests green on <base>@<sha>; migrated <test, staging>.
Evidence: <the gate's `needs-you:` lines, verbatim>
Reason tag: 🙋 needs you · Owner: <Eric | repo admin>
Reason tag: 🙋 needs you · Owner: <owner | repo admin>
approval-request: <copy the gate's exact JSON here>
blocked-by: -
</details>
Expand Down
2 changes: 1 addition & 1 deletion skills/super-board/references/run-workflow.md
Original file line number Diff line number Diff line change
Expand Up @@ -278,7 +278,7 @@ Repeat until a done condition or halt gate fires:

```
Wave 3 · 5 cards → 2 merged · 1 bounced · 1 blocked · 1 failed · usage 5h 62%
🛡 #15 Blocked 🔐 missing STRIPE_TEST_KEY — owner: Eric
🛡 #15 Blocked 🔐 missing STRIPE_TEST_KEY — needs owner
❌ #18 failed: build lane returned no result — retried next wave
Next: wave 4 (3 cards) | ⏸ paused until 14:05 | ✅ board drained
```
Expand Down
4 changes: 2 additions & 2 deletions skills/super-board/references/run.md
Original file line number Diff line number Diff line change
Expand Up @@ -289,7 +289,7 @@ The agent may turn a `proceed` into a `hold` on evidence. It may turn a mechanic
| Verdict | Action | Block template (`block-template.md`) |
|---|---|---|
| `proceed` | nothing; the Builder starts | — |
| `hold` — done | move card to Blocked | `👯` · names the merged PR / closed issue · Owner: Eric (close as duplicate) · `blocked-by: -` |
| `hold` — done | move card to Blocked | `👯` · names the merged PR / closed issue · Owner: owner (close as duplicate) · `blocked-by: -` |
| `hold` — in progress | move card to Blocked | `👯` · names the open PR / card · `blocked-by: <the issue that PR or card closes>` — the sweep returns it when that closes, and pre-flight then finds it done |
| `hold` — unclear | move card to Blocked | `❓` · the missing AC or the contradiction with `file:line` · `blocked-by: -` |
| `hold` — too big | move card to Blocked | `❓` · "too big — likely over <cap> changed lines / spans <areas>" · suggests splitting with `/to-tickets` into vertical slices, each under the cap · `blocked-by: -` |
Expand Down Expand Up @@ -523,7 +523,7 @@ Every lane writes BOTH on every exit, in the comment format of writing-standard.
Did: <one line>
✅ Done: <what passed or landed, with sha / file:line / command>
❌ Not done: <what failed or was skipped, and why> (omit when nothing)
Next: <builder | qa | reviewer | Eric | none>
Next: <builder | qa | reviewer | owner | none>
```

- Role = who writes it: `builder` · `qa` · `reviewer` · `collect` · `orchestrator`. Label for a
Expand Down
Loading