mikeparcewski/wicked-garden / skills/draft
wicked-garden-draft
VerifiedThe quality floor for DOCUMENT deliverables drafted from a repo — brochures, one-pagers, flyers, pages, decks, memos — guidance plus a self-check before "done". Three floors: (1) READABLE — every text/background pair ≥ 4.5:1 (WCAG AA; 3:1 for large text) on COMPOSITED colours, print text ≥ 7pt; (2) SIZED TO THE BRIEF — the RENDERED page count honours the brief's page budget; (3) GROUNDED — every number, URL and claim cites a repo path, mocks visibly labelled, no placeholder ships. Use when: "draft a brochure / one-pager / page / deck / memo about <repo>", "print-ready A4 / Letter", "make it two pages", "is this brochure readable / grounded", before declaring a wicked-interactive draft/edit/revision done, or as the skill a governed drafting run (wicked-crew `interactive-*`) follows. A brief's `Kind:` line (narrative, status update, message, …) adds that kind's ref. NOT for code documentation (the engineering documentarian) or a running app's accessibility audit (the product or qe a11y skills).
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 14.0 KB |
| refs/claims.md | markdown | 4.7 KB |
| refs/demo-script.md | markdown | 1.9 KB |
| refs/design-floor.md | markdown | 4.8 KB |
| refs/meeting-prep.md | markdown | 1.9 KB |
| refs/message.md | markdown | 1.6 KB |
| refs/narrative.md | markdown | 1.9 KB |
| refs/operating-model.md | markdown | 2.1 KB |
| refs/respond.md | markdown | 3.9 KB |
| refs/rework.md | markdown | 4.1 KB |
| refs/self-check.md | markdown | 7.0 KB |
| refs/status-update.md | markdown | 1.7 KB |
refs/claims.md
View on GitHub# Claims discipline — the citation convention
The grounded floor of the `wicked-garden-draft` skill. Re-derived by `scripts/draft/claims_scan.py`
over the VISIBLE text of the deliverable.
## What is a claim
Any statement a reader could check against the repository: a version or requirement
(`Node ≥ 22`), a number (`within 2 s`, `94 %`, `3 active`), a URL or domain, a quote, a named
capability ("desktop notifications when a run needs you"), a licence, a supported platform.
The scan judges numbers and URLs mechanically; you judge the rest by the same rule.
## How to cite (any ONE of these makes a block "cited")
1. **`data-source` on the element or its block** — the primary form; invisible in print and
safe under wicked-interactive's instrumentation (only `data-wid` is theirs):
```html
<li data-source="README.md:23-24">Human gates: approve · steer · reject</li>
<section class="page page-2" data-source=".product/REQ-001-application-overview.md:82 README.md:60">
```
Several sources: separate with spaces or commas. `path[:line[-line]]`, repo-relative.
2. **An inline repo path in the same block** — `Requires Node ≥ 22 (package.json:72)`.
3. **A footnote marker (`[3]`, `†`) AND a visible Sources block** — a `Sources` / `References`
heading (or an element with `data-sources`) followed by the repo paths.
## What a citation is verified against (the rule)
A `data-source="path:N"` asserts **the text of the element that carries it**, so the scan verifies
it against that element's own text — its inline children included, a sibling's text never:
| Where the citation sits | Verified against |
|---|---|
| the element with the text | that element's text |
| an inline wrapper (`<b>`, `<span>`) around it | the wrapper's text |
| the enclosing block (`<li>`, `<p>`) | the whole block's text |
| an ancestor above the block (`<section>`) | nothing — inherited, path existence only |
| path-only, no `:N` | nothing — provenance, path existence only |
Two `<span>`s in one `<li>`, each with its own `data-source`, are therefore two separate claims:
the first one's text cannot satisfy the second one's citation. Put the citation on the element
whose text it proves.
A citation passes only when the cited text shares a **content** word with the claim — an overlap
of common words alone ("the", "is", "under") verifies nothing and is reported. A range
(`path:23-24`) is judged against the best span of its own width, so a wide range cannot tie its
way to a pass; cite the narrowest span that carries the text, and expect ranges to be counted
separately in the summary ("N by line range") because they assert less than a line does.
Give the reader a visible **Sources** strip as well (form 3's block, or a footer line listing
the paths): the invisible attribute proves the claim; the strip lets a human check it.
## URLs and domains
- Every `https://…` and every bare domain (`ws.wickedagile.com`) is a claim. With `--repo`,
the scan requires the address to appear in some file of the snapshot (`unsourced-url`
otherwise); without a repo it requires a `data-source` (`uncited-url`).
- Existence is not meaning. RAID.md documents `ws.wickedagile.com` as the **marketing /
deep-dive site**; calling it "the live wicked-studio instance" is an invented claim the scan
cannot see — read the source line and characterise the link the way the source does.
## Mocks, illustrations, sample data
A hero "command center" with `impl/auth-layer running · 3 active · 1 needs you` is fine ONLY
when the reader is told it is not real:
```html
<div class="hero-mockup" data-illustrative>
<div class="mockup-caption">Illustrative — sample runs, not measured data</div>
…
</div>
```
- `data-illustrative` (or `data-mock`) on the container, AND a visible label inside it
containing "illustrative", "example", "mock", "sample" or "hypothetical".
- A label that lives only in an HTML comment is the F-RECON-009 defect: `mock-label-hidden`.
- A container with the attribute and no visible label: `mock-unlabelled`.
## Placeholders
Never in the deliverable. `[PLACEHOLDER: …]`, `TODO`, `TBD`, `FIXME`, `lorem ipsum`,
`{{variable}}`, `[INSERT …]`, `[Company name here]` are all `placeholder` findings when visible.
The gap goes in your reply/notes: *"No customer quote or third-party audit exists in the
repository — the validation section was omitted rather than filled with a placeholder."*
## What the scan deliberately ignores
- Section ordinals written with a leading zero (`01`, `02`), copyright years (`© 2026`).
- Numbers inside a cited path or URL (`server.ts:1097`, `/v2/docs`).
- Text in comments, `<script>`, `<style>`, `hidden` elements — a reader never sees it.
- Prose claims without numbers or URLs (the model's job: cite them the same way).
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-analysis-optimizationmorluto/reaOptimize REA providers and shared analysis workflows for agent context, latency, CPU, and memory while preserving correctness and Evidence. Use for analysis bottlenecks and cache or query design; CI/test-suite optimization belongs to rea-ci-tests-audit. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-ci-tests-auditmorluto/reaAudit REA CI and test suites for correctness, fixture drift, false passes, and measured performance; use for suite or workflow investigations and efficiency changes, rather than ordinary feature testing or a single test failure. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| rea-tool-designmorluto/reaAudit, investigate, design, or change REA investigation tools, CLI/MCP contracts, provider capabilities, and Evidence semantics. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| ponytail-auditDietrichGebert/ponytailQuality audit of the whole repo: bugs, security, real load, missing tests, speed, and what to delete. Most important first. | Stars: 194k | Updated | Claude CodeCodex | Verified | Install |