[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M77M3JPRMDMZ9AZ9ECVG3FMA":3,"skill-same:01M77M3JPRMDMZ9AZ9ECVG3FMA":55},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M77M3JPRMDMZ9AZ9ECVG3FMA","spec-visualization",null,"Produce a single-file interactive HTML dashboard visualizing a finalized spec — architectural planes, end-to-end pipeline (React Flow), wave timeline, review evolution, decisions, components, open gates. Self-contained, opens in browser, no build step. Trigger after spec-review fixes are applied, or on demand for any approved spec.","Roeelfs\u002Fkeel","Roeelfs","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F77358299?v=4","01M77KYYCM1M5SB4RSAD5NKZWH","Roee Alfasi","01M77KYYDD96C6P8AT0ZF0VTYY",1,"MIT","https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fspec-visualization\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nlicense: MIT\nname: spec-visualization\ndescription: Produce a single-file interactive HTML dashboard visualizing a finalized spec — architectural planes, end-to-end pipeline (React Flow), wave timeline, review evolution, decisions, components, open gates. Self-contained, opens in browser, no build step. Trigger after spec-review fixes are applied, or on demand for any approved spec.\n---\n\n# Spec Visualization\n\nProduces a single HTML file that lets a reader grok an entire spec at a glance: the 4-plane data-boundary architecture, the commit→deploy pipeline as an interactive flow chart, wave-by-wave rollout with gating items, review evolution charts, decisions status, component inventory, open items dashboard.\n\n## When to invoke\n\n- **As Step 10 of `spec-review`** — after fixes are applied, before commit. Skip if spec is still Draft or no review files exist.\n- **Standalone** — when the user asks to \"visualize the spec\", \"see the spec model\", \"render the design\", or wants a fitness check against a vision.\n- **NOT for** specs under 200 lines (overkill) or specs with no waves\u002Freviews (the dashboard sections won't have data).\n\n## Output\n\n`\u003Cspec-path>.viz.html` — single HTML file next to the spec. Opens in Chrome. Self-contained (CDN imports for React 18, @xyflow\u002Freact, Recharts, Tailwind, lucide-react, Babel standalone).\n\n## Process\n\n### Step 1: Locate inputs\n\n- **Spec file** — the canonical .md (typically `docs\u002Fspecs\u002Factive\u002FYYYY-MM-DD-\u003Cname>.md`)\n- **Review files** — any sibling `\u003Cspec-basename>.review*.md` files\n- **Decisions doc** — sibling `*-decisions.md` if present\n- **Related platform specs** — sibling specs referenced in the spec's frontmatter\n\nIf only the spec exists with no reviews\u002Fdecisions, the dashboard renders with those sections empty — that's fine.\n\n### Step 2: Extract data into the canonical schema\n\nRead the spec end-to-end. Map its content onto this canonical schema (the full field reference is defined inline below):\n\n```javascript\n{\n  \"meta\": {\n    \"title\": \"...\",                          \u002F\u002F from spec H1\n    \"specPath\": \"docs\u002Fspecs\u002Factive\u002F...\",\n    \"version\": \"v0.3.1\",                     \u002F\u002F from \"Status:\" line\n    \"issueRef\": \"#379\",                      \u002F\u002F from \"Issue:\" line\n    \"branch\": \"spec\u002F...\",                    \u002F\u002F git branch the spec lives on\n    \"statusBadges\": [                        \u002F\u002F array of {label, variant}\n      {\"label\": \"v0.3.1\", \"variant\": \"accent\"},\n      {\"label\": \"Wave 1 ready\", \"variant\": \"success\"}\n    ],\n    \"summary\": \"...\"                         \u002F\u002F 1-2 sentence top-of-page summary\n  },\n  \"planes\": [                                \u002F\u002F architectural layers, typically 3-5\n    {\n      \"id\": \"workspace\",\n      \"label\": \"Plane 1 · Authoring\",\n      \"title\": \"Workspace\",\n      \"store\": \"Git repo · per project\",\n      \"desc\": \"Canonical multi-file source...\",\n      \"items\": [\"manifest.json\", \"schema.json\", \"...\"],\n      \"color\": \"violet\",                     \u002F\u002F violet | indigo | cyan | emerald | amber | rose\n      \"icon\": \"GitBranch\"                    \u002F\u002F lucide-react icon name\n    }\n  ],\n  \"pipelineNodes\": [                         \u002F\u002F React Flow nodes\n    {\n      \"id\": \"agent\",\n      \"position\": {\"x\": 0, \"y\": 0},\n      \"data\": {\n        \"label\": \"Agent\",\n        \"sub\": \"build agent\",\n        \"plane\": \"workspace\",                \u002F\u002F matches a plane id\n        \"icon\": \"Sparkles\"\n      }\n    }\n  ],\n  \"pipelineEdges\": [                         \u002F\u002F React Flow edges\n    {\n      \"id\": \"e1\",\n      \"source\": \"agent\",\n      \"target\": \"validate\",\n      \"label\": \"pushed\",                     \u002F\u002F optional\n      \"animated\": true,                      \u002F\u002F optional\n      \"color\": \"violet\",                     \u002F\u002F matches plane palette\n      \"dashed\": false                        \u002F\u002F optional\n    }\n  ],\n  \"reviewRounds\": [                          \u002F\u002F chronological\n    {\"round\": \"v0.1\", \"critical\": 7, \"major\": 14, \"minor\": 9, \"verdict\": \"block\"},\n    {\"round\": \"v0.2\", \"critical\": 10, \"major\": 18, \"minor\": 13, \"verdict\": \"block\"}\n  ],\n  \"waves\": [                                 \u002F\u002F rollout plan, typically from §17\n    {\n      \"id\": 0,\n      \"title\": \"Spec + Model\",\n      \"state\": \"done\",                       \u002F\u002F done | next | pending\n      \"ships\": [\"Spec finalized\", \"...\"],\n      \"gates\": [\n        {\"id\": \"NEW-NC7\", \"sev\": \"critical\", \"label\": \"I-1 amendment\"},\n        {\"id\": \"NEW-M2\",  \"sev\": \"fixed\",    \"label\": \"AND→OR\"}\n      ]\n    }\n  ],\n  \"decisions\": [                             \u002F\u002F Wave 0 \u002F open decisions\n    {\n      \"id\": \"D1\",\n      \"topic\": \"Wave 5\u002F6 ordering\",\n      \"flag\": \"NEW-NC8\",\n      \"rec\": \"Merge Wave 6 INTO Wave 5...\",\n      \"owner\": \"platform-architecture\",\n      \"status\": \"pending\",                   \u002F\u002F pending | updated | approved\n      \"note\": null                           \u002F\u002F optional expanded note\n    }\n  ],\n  \"components\": {                            \u002F\u002F infra inventory grouped\n    \"db\": {\n      \"label\": \"Database tables\",\n      \"count\": \"8 + 1 queue\",\n      \"icon\": \"Database\",\n      \"items\": [\n        {\"name\": \"project_repositories\", \"note\": \"one row per project · pointers + state\"}\n      ]\n    }\n  },\n  \"openItemsByWave\": [                       \u002F\u002F for stacked bar\n    {\"wave\": \"W0\", \"critical\": 1, \"major\": 0}\n  ],\n  \"boundaries\": [                            \u002F\u002F data-store boundaries narrative\n    {\n      \"color\": \"violet\",\n      \"label\": \"Git repo \u002F object store\",\n      \"title\": \"Configuration\",\n      \"contents\": \"schema.json · automations · ...\"\n    }\n  ]\n}\n```\n\n**Extraction heuristics:**\n\n- **Planes** — most specs have a \"§4.1 Planes\" or \"Architecture\" section with a table. If not, infer from the data-boundaries section. Typical specs have 3-4 planes mapping to data stores.\n- **Pipeline nodes** — read the \"High-Level Flow\" \u002F sequence diagram \u002F §X.Y materializer narrative. Each step becomes a node. Aim for 8-18 nodes — fewer than 8 is too sparse, more than 18 is overwhelming.\n- **Pipeline edges** — connect nodes in flow order. Color edges by source plane. Use `animated: true` for the critical happy path; static for diverging branches. Add labels for state transitions (e.g., `pushed`, `bundled`, `accepted`).\n\n- **Pipeline layout — strict lane discipline (do this, not ad-hoc x\u002Fy).** Overlapping edges are the #1 readability failure. React Flow uses `smoothstep` (orthogonal) routing, so the layout must be planned as horizontal lanes:\n  - **Stride:** `x` += **230** per pipeline stage (nodes are ~140-200px wide; 230 leaves a clean gap). `y` lanes **≥130px apart**.\n  - **The spine is sacred.** The happy path lives on **one lane, `y: 0`**, strictly left-to-right. Nothing else may occupy `y: 0`, and **no edge may route through the `y: 0` band between two spine nodes** except the spine edge itself.\n  - **Branches get their own lanes, never the spine.** Parallel\u002Falternative concerns (e.g. triage axes) → a vertical **fan column** at one shared `x`, stacked at `y: -180, -50, 80, 210`. Recovery \u002F safety \u002F out-of-band → **lower lanes** (`y: 300+`), one lane per concern, running *parallel* to the spine then turning up into their join node — never diagonally across it.\n  - **Order branch endpoints to match.** When several edges fan from one node to a column (or converge from a column into one node), order the targets vertically in the same order the edges leave\u002Fenter so lines don't cross each other.\n  - **Feedback \u002F backward edges** (loops returning to an earlier stage) go on the **lowest empty lane**, `dashed: true`, distinct color, with a label. Accept one clean long edge there; never route a backward edge through the spine or node bands.\n  - **Long-distance edges hug a lane.** An edge spanning many stages must travel along an empty `y` lane (its own), not cut diagonally. If two long edges would share a lane, give one a lane 130px lower.\n  - **Collapse before you lay out.** Template nodes have a *single* left (target) \u002F right (source) handle, so every off-spine edge loops around — N parallel sub-concerns rendered as N nodes produces N looping edges and guaranteed overlap. Model parallel\u002Falternative concerns as **one spine node** with a summarizing `sub` (e.g. \"Three-axis triage · urgency · safety · statutory\"); push the per-item detail into the Components section, not the flow. A flow chart shows *flow*, not every rule. Aim for a near-linear spine with ≤4 short off-spine edges + ≤2 isolated dashed lanes (one branch, one feedback).\n  - Sanity check before render: every edge is either (a) a short spine hop, (b) a short confirm→{outputs,followup}-style stub, or (c) one of the ≤2 lane-hugging branch\u002Ffeedback edges. Any edge that visually crosses a node = collapse nodes or relayout.\n- **Review rounds** — count CRITICAL\u002FMAJOR\u002FMINOR per `.review*.md` file. The spec usually has a \"review history\" or change-log section with these counts.\n- **Waves** — from §17 (or \"Rollout Plan\"). The \"Must close before exit\" subsections are the gates. Mark wave state: `done` (status indicates done), `next` (the immediately-next wave), `pending` (everything later).\n- **Decisions** — from any `*-decisions.md` companion. If none, omit the section.\n- **Components** — from the spec's infrastructure sections: database tables, config\u002Fcredential stores, inferred compute units (functions\u002Fservices) from handler refs, encryption keys from the secrets section.\n- **Open items by wave** — count gating items in each wave by severity. Used for the stacked bar chart.\n\nIf the spec is too unstructured to extract any of these, drop the section from the data — the template renders gracefully with missing fields.\n\n### Step 3: Render\n\nCopy `template.html` from this skill directory to `\u003Cspec-path>.viz.html`. Locate the `\u003Cscript id=\"spec-data\" type=\"application\u002Fjson\">` block in the template and replace its inner text with the JSON-stringified data object from Step 2.\n\n```bash\nTEMPLATE=\"$HOME\u002F.claude\u002Fskills\u002Fspec-visualization\u002Ftemplate.html\"\nOUTPUT=\"${SPEC_PATH%.md}.viz.html\"\n\n# Read template, inject JSON data, write output. Use Python to avoid shell-escaping JSON:\npython3 - \"$TEMPLATE\" \"$OUTPUT\" \"$DATA_JSON_PATH\" \u003C\u003C'PYEOF'\nimport sys, json, re\ntemplate = open(sys.argv[1]).read()\ndata = json.load(open(sys.argv[3]))\ndata_str = json.dumps(data, indent=2)\nout = re.sub(\n    r'\u003Cscript id=\"spec-data\" type=\"application\u002Fjson\">[\\s\\S]*?\u003C\u002Fscript>',\n    f'\u003Cscript id=\"spec-data\" type=\"application\u002Fjson\">{data_str}\u003C\u002Fscript>',\n    template, count=1\n)\nopen(sys.argv[2], 'w').write(out)\nPYEOF\n```\n\n### Step 4: Open + report\n\n```bash\nopen -a \"Google Chrome\" \"$OUTPUT\"\n```\n\nTell the user: file path, what sections rendered, any sections skipped (and why).\n\n### Step 5: Commit (optional)\n\nIf the spec is on a branch, offer to commit the `.viz.html` next to the spec:\n\n```bash\ngit add \"$OUTPUT\"\ngit commit -m \"docs(\u003Cscope>): visualize spec — interactive dashboard\"\n```\n\nThe `.viz.html` is reproducible from the spec, so committing is optional. If the spec is volatile (Draft early), don't commit.\n\n## Tips\n\n- **Layout the pipeline once** — apply the strict lane discipline from the extraction heuristics (spine at `y:0`, 230 x-stride, branches\u002Ffeedback on their own lower lanes). Re-renders preserve positions; record final `x`\u002F`y` in the data. If any edge visually crosses a node after the first render, relayout — don't ship overlapping edges.\n- **Wave count >8** — switch the wave timeline from horizontal to vertical lanes. Edit the template's wave-grid Tailwind class from `lg:grid-cols-8` to `lg:grid-cols-4` for two rows.\n- **No review history** — drop the review evolution section entirely. The page is still informative without it.\n- **Light mode** — the template is dark-only by design (operational dashboard aesthetic). Light mode adds complexity without payoff.\n- **Failure paths** — color them rose. Use `dashed: true` on edges that represent recovery \u002F out-of-band paths.\n\n## When NOT to use\n\n- Specs without a clear architectural model (e.g., bug fixes, refactor specs) — the planes\u002Fpipeline sections won't have meaningful content\n- One-shot reference docs (CLAUDE.md, runbooks) — those aren't specs\n- Pre-Draft specs — the dashboard implies stability the spec hasn't reached\n\n## LEARNINGS.md\n\nAppend findings under **What Worked** \u002F **Patterns** as you go. Specific entries worth saving: which extraction heuristic worked for a non-obvious spec shape, what makes a pipeline diagram readable, when to drop a section.\n","# Spec Visualization\n\nProduces a single HTML file that lets a reader grok an entire spec at a glance: the 4-plane data-boundary architecture, the commit→deploy pipeline as an interactive flow chart, wave-by-wave rollout with gating items, review evolution charts, decisions status, component inventory, open items dashboard.\n\n## When to invoke\n\n- **As Step 10 of `spec-review`** — after fixes are applied, before commit. Skip if spec is still Draft or no review files exist.\n- **Standalone** — when the user asks to \"visualize the spec\", \"see the spec model\", \"render the design\", or wants a fitness check against a vision.\n- **NOT for** specs under 200 lines (overkill) or specs with no waves\u002Freviews (the dashboard sections won't have data).\n\n## Output\n\n`\u003Cspec-path>.viz.html` — single HTML file next to the spec. Opens in Chrome. Self-contained (CDN imports for React 18, @xyflow\u002Freact, Recharts, Tailwind, lucide-react, Babel standalone).\n\n## Process\n\n### Step 1: Locate inputs\n\n- **Spec file** — the canonical .md (typically `docs\u002Fspecs\u002Factive\u002FYYYY-MM-DD-\u003Cname>.md`)\n- **Review files** — any sibling `\u003Cspec-basename>.review*.md` files\n- **Decisions doc** — sibling `*-decisions.md` if present\n- **Related platform specs** — sibling specs referenced in the spec's frontmatter\n\nIf only the spec exists with no reviews\u002Fdecisions, the dashboard renders with those sections empty — that's fine.\n\n### Step 2: Extract data into the canonical schema\n\nRead the spec end-to-end. Map its content onto this canonical schema (the full field reference is defined inline below):\n\n```javascript\n{\n  \"meta\": {\n    \"title\": \"...\",                          \u002F\u002F from spec H1\n    \"specPath\": \"docs\u002Fspecs\u002Factive\u002F...\",\n    \"version\": \"v0.3.1\",                     \u002F\u002F from \"Status:\" line\n    \"issueRef\": \"#379\",                      \u002F\u002F from \"Issue:\" line\n    \"branch\": \"spec\u002F...\",                    \u002F\u002F git branch the spec lives on\n    \"statusBadges\": [                        \u002F\u002F array of {label, variant}\n      {\"label\": \"v0.3.1\", \"variant\": \"accent\"},\n      {\"label\": \"Wave 1 ready\", \"variant\": \"success\"}\n    ],\n    \"summary\": \"...\"                         \u002F\u002F 1-2 sentence top-of-page summary\n  },\n  \"planes\": [                                \u002F\u002F architectural layers, typically 3-5\n    {\n      \"id\": \"workspace\",\n      \"label\": \"Plane 1 · Authoring\",\n      \"title\": \"Workspace\",\n      \"store\": \"Git repo · per project\",\n      \"desc\": \"Canonical multi-file source...\",\n      \"items\": [\"manifest.json\", \"schema.json\", \"...\"],\n      \"color\": \"violet\",                     \u002F\u002F violet | indigo | cyan | emerald | amber | rose\n      \"icon\": \"GitBranch\"                    \u002F\u002F lucide-react icon name\n    }\n  ],\n  \"pipelineNodes\": [                         \u002F\u002F React Flow nodes\n    {\n      \"id\": \"agent\",\n      \"position\": {\"x\": 0, \"y\": 0},\n      \"data\": {\n        \"label\": \"Agent\",\n        \"sub\": \"build agent\",\n        \"plane\": \"workspace\",                \u002F\u002F matches a plane id\n        \"icon\": \"Sparkles\"\n      }\n    }\n  ],\n  \"pipelineEdges\": [                         \u002F\u002F React Flow edges\n    {\n      \"id\": \"e1\",\n      \"source\": \"agent\",\n      \"target\": \"validate\",\n      \"label\": \"pushed\",                     \u002F\u002F optional\n      \"animated\": true,                      \u002F\u002F optional\n      \"color\": \"violet\",                     \u002F\u002F matches plane palette\n      \"dashed\": false                        \u002F\u002F optional\n    }\n  ],\n  \"reviewRounds\": [                          \u002F\u002F chronological\n    {\"round\": \"v0.1\", \"critical\": 7, \"major\": 14, \"minor\": 9, \"verdict\": \"block\"},\n    {\"round\": \"v0.2\", \"critical\": 10, \"major\": 18, \"minor\": 13, \"verdict\": \"block\"}\n  ],\n  \"waves\": [                                 \u002F\u002F rollout plan, typically from §17\n    {\n      \"id\": 0,\n      \"title\": \"Spec + Model\",\n      \"state\": \"done\",                       \u002F\u002F done | next | pending\n      \"ships\": [\"Spec finalized\", \"...\"],\n      \"gates\": [\n        {\"id\": \"NEW-NC7\", \"sev\": \"critical\", \"label\": \"I-1 amendment\"},\n        {\"id\": \"NEW-M2\",  \"sev\": \"fixed\",    \"label\": \"AND→OR\"}\n      ]\n    }\n  ],\n  \"decisions\": [                             \u002F\u002F Wave 0 \u002F open decisions\n    {\n      \"id\": \"D1\",\n      \"topic\": \"Wave 5\u002F6 ordering\",\n      \"flag\": \"NEW-NC8\",\n      \"rec\": \"Merge Wave 6 INTO Wave 5...\",\n      \"owner\": \"platform-architecture\",\n      \"status\": \"pending\",                   \u002F\u002F pending | updated | approved\n      \"note\": null                           \u002F\u002F optional expanded note\n    }\n  ],\n  \"components\": {                            \u002F\u002F infra inventory grouped\n    \"db\": {\n      \"label\": \"Database tables\",\n      \"count\": \"8 + 1 queue\",\n      \"icon\": \"Database\",\n      \"items\": [\n        {\"name\": \"project_repositories\", \"note\": \"one row per project · pointers + state\"}\n      ]\n    }\n  },\n  \"openItemsByWave\": [                       \u002F\u002F for stacked bar\n    {\"wave\": \"W0\", \"critical\": 1, \"major\": 0}\n  ],\n  \"boundaries\": [                            \u002F\u002F data-store boundaries narrative\n    {\n      \"color\": \"violet\",\n      \"label\": \"Git repo \u002F object store\",\n      \"title\": \"Configuration\",\n      \"contents\": \"schema.json · automations · ...\"\n    }\n  ]\n}\n```\n\n**Extraction heuristics:**\n\n- **Planes** — most specs have a \"§4.1 Planes\" or \"Architecture\" section with a table. If not, infer from the data-boundaries section. Typical specs have 3-4 planes mapping to data stores.\n- **Pipeline nodes** — read the \"High-Level Flow\" \u002F sequence diagram \u002F §X.Y materializer narrative. Each step becomes a node. Aim for 8-18 nodes — fewer than 8 is too sparse, more than 18 is overwhelming.\n- **Pipeline edges** — connect nodes in flow order. Color edges by source plane. Use `animated: true` for the critical happy path; static for diverging branches. Add labels for state transitions (e.g., `pushed`, `bundled`, `accepted`).\n\n- **Pipeline layout — strict lane discipline (do this, not ad-hoc x\u002Fy).** Overlapping edges are the #1 readability failure. React Flow uses `smoothstep` (orthogonal) routing, so the layout must be planned as horizontal lanes:\n  - **Stride:** `x` += **230** per pipeline stage (nodes are ~140-200px wide; 230 leaves a clean gap). `y` lanes **≥130px apart**.\n  - **The spine is sacred.** The happy path lives on **one lane, `y: 0`**, strictly left-to-right. Nothing else may occupy `y: 0`, and **no edge may route through the `y: 0` band between two spine nodes** except the spine edge itself.\n  - **Branches get their own lanes, never the spine.** Parallel\u002Falternative concerns (e.g. triage axes) → a vertical **fan column** at one shared `x`, stacked at `y: -180, -50, 80, 210`. Recovery \u002F safety \u002F out-of-band → **lower lanes** (`y: 300+`), one lane per concern, running *parallel* to the spine then turning up into their join node — never diagonally across it.\n  - **Order branch endpoints to match.** When several edges fan from one node to a column (or converge from a column into one node), order the targets vertically in the same order the edges leave\u002Fenter so lines don't cross each other.\n  - **Feedback \u002F backward edges** (loops returning to an earlier stage) go on the **lowest empty lane**, `dashed: true`, distinct color, with a label. Accept one clean long edge there; never route a backward edge through the spine or node bands.\n  - **Long-distance edges hug a lane.** An edge spanning many stages must travel along an empty `y` lane (its own), not cut diagonally. If two long edges would share a lane, give one a lane 130px lower.\n  - **Collapse before you lay out.** Template nodes have a *single* left (target) \u002F right (source) handle, so every off-spine edge loops around — N parallel sub-concerns rendered as N nodes produces N looping edges and guaranteed overlap. Model parallel\u002Falternative concerns as **one spine node** with a summarizing `sub` (e.g. \"Three-axis triage · urgency · safety · statutory\"); push the per-item detail into the Components section, not the flow. A flow chart shows *flow*, not every rule. Aim for a near-linear spine with ≤4 short off-spine edges + ≤2 isolated dashed lanes (one branch, one feedback).\n  - Sanity check before render: every edge is either (a) a short spine hop, (b) a short confirm→{outputs,followup}-style stub, or (c) one of the ≤2 lane-hugging branch\u002Ffeedback edges. Any edge that visually crosses a node = collapse nodes or relayout.\n- **Review rounds** — count CRITICAL\u002FMAJOR\u002FMINOR per `.review*.md` file. The spec usually has a \"review history\" or change-log section with these counts.\n- **Waves** — from §17 (or \"Rollout Plan\"). The \"Must close before exit\" subsections are the gates. Mark wave state: `done` (status indicates done), `next` (the immediately-next wave), `pending` (everything later).\n- **Decisions** — from any `*-decisions.md` companion. If none, omit the section.\n- **Components** — from the spec's infrastructure sections: database tables, config\u002Fcredential stores, inferred compute units (functions\u002Fservices) from handler refs, encryption keys from the secrets section.\n- **Open items by wave** — count gating items in each wave by severity. Used for the stacked bar chart.\n\nIf the spec is too unstructured to extract any of these, drop the section from the data — the template renders gracefully with missing fields.\n\n### Step 3: Render\n\nCopy `template.html` from this skill directory to `\u003Cspec-path>.viz.html`. Locate the `\u003Cscript id=\"spec-data\" type=\"application\u002Fjson\">` block in the template and replace its inner text with the JSON-stringified data object from Step 2.\n\n```bash\nTEMPLATE=\"$HOME\u002F.claude\u002Fskills\u002Fspec-visualization\u002Ftemplate.html\"\nOUTPUT=\"${SPEC_PATH%.md}.viz.html\"\n\n# Read template, inject JSON data, write output. Use Python to avoid shell-escaping JSON:\npython3 - \"$TEMPLATE\" \"$OUTPUT\" \"$DATA_JSON_PATH\" \u003C\u003C'PYEOF'\nimport sys, json, re\ntemplate = open(sys.argv[1]).read()\ndata = json.load(open(sys.argv[3]))\ndata_str = json.dumps(data, indent=2)\nout = re.sub(\n    r'\u003Cscript id=\"spec-data\" type=\"application\u002Fjson\">[\\s\\S]*?\u003C\u002Fscript>',\n    f'\u003Cscript id=\"spec-data\" type=\"application\u002Fjson\">{data_str}\u003C\u002Fscript>',\n    template, count=1\n)\nopen(sys.argv[2], 'w').write(out)\nPYEOF\n```\n\n### Step 4: Open + report\n\n```bash\nopen -a \"Google Chrome\" \"$OUTPUT\"\n```\n\nTell the user: file path, what sections rendered, any sections skipped (and why).\n\n### Step 5: Commit (optional)\n\nIf the spec is on a branch, offer to commit the `.viz.html` next to the spec:\n\n```bash\ngit add \"$OUTPUT\"\ngit commit -m \"docs(\u003Cscope>): visualize spec — interactive dashboard\"\n```\n\nThe `.viz.html` is reproducible from the spec, so committing is optional. If the spec is volatile (Draft early), don't commit.\n\n## Tips\n\n- **Layout the pipeline once** — apply the strict lane discipline from the extraction heuristics (spine at `y:0`, 230 x-stride, branches\u002Ffeedback on their own lower lanes). Re-renders preserve positions; record final `x`\u002F`y` in the data. If any edge visually crosses a node after the first render, relayout — don't ship overlapping edges.\n- **Wave count >8** — switch the wave timeline from horizontal to vertical lanes. Edit the template's wave-grid Tailwind class from `lg:grid-cols-8` to `lg:grid-cols-4` for two rows.\n- **No review history** — drop the review evolution section entirely. The page is still informative without it.\n- **Light mode** — the template is dark-only by design (operational dashboard aesthetic). Light mode adds complexity without payoff.\n- **Failure paths** — color them rose. Use `dashed: true` on edges that represent recovery \u002F out-of-band paths.\n\n## When NOT to use\n\n- Specs without a clear architectural model (e.g., bug fixes, refactor specs) — the planes\u002Fpipeline sections won't have meaningful content\n- One-shot reference docs (CLAUDE.md, runbooks) — those aren't specs\n- Pre-Draft specs — the dashboard implies stability the spec hasn't reached\n\n## LEARNINGS.md\n\nAppend findings under **What Worked** \u002F **Patterns** as you go. Specific entries worth saving: which extraction heuristic worked for a non-obvious spec shape, what makes a pipeline diagram readable, when to drop a section.\n",{"name":5,"license":15,"description":7},"master","45893bf9b7c9a6a59fc8fb0be2dee4a6fec11811",".claude\u002Fskills\u002Fspec-visualization",0,"https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel","https:\u002F\u002Fgithub.com\u002FRoeelfs","2026-11-13 18:38:08",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add Roeelfs\u002Fkeel --skill spec-visualization",{"tool":37,"command":38},"bunx","bunx skills add Roeelfs\u002Fkeel --skill spec-visualization",{"tool":40,"command":41},"pnpm","pnpm dlx skills add Roeelfs\u002Fkeel --skill spec-visualization","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fspec-visualization\u002FSKILL.md",true,[45,51],{"path":46,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":49,"sourceHtmlUrl":50},"LEARNINGS.md","REFERENCE","参考文档",1451,"https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fspec-visualization\u002FLEARNINGS.md",{"path":52,"resourceType":47,"resourceTypeDesc":48,"sizeBytes":53,"sourceHtmlUrl":54},"template.html",28608,"https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fspec-visualization\u002Ftemplate.html",[56,63,70,77,84,91],{"skillNo":57,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M77M4PVP0XRR0QG4WVY67MXM","writing-for-agents","Writing documents for agents. Use when creating or editing skills, or modifying AGENTS.md or CLAUDE.md.","https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fwriting-for-agents\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M77M4K4HQTVPC61S4KQY7K71","work-report","Use when generating development reports, calculating billable hours, reviewing work done in a period, or when the user asks for a project status summary. Triggers on keywords like report, billable, hours, invoice, timesheet, work summary.","https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fwork-report\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M77M4G8XZ87W5KE13E43GS54","wizard","Generate an interactive bash wizard that walks a human through steps only they can perform. Use when provisioning infrastructure, setting up credentials or CI secrets, walking an unfamiliar third-party dashboard, or running a one-off migration or cutover. Don't invoke this for steps the agent can perform itself.","https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fwizard\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M77M4CSCMXVBGN4DYK8PZXW2","wayfinder","Plan a huge chunk of work (more than one agent session can hold) as a shared map of decision tickets on your issue tracker, and resolve them one at a time until the way to the destination is clear.","https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fwayfinder\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M77M4A82NCN0FP3W1SKBFZAD","triage","Move issues and external PRs through a state machine of triage roles, categorise, verify, grill if needed, and write agent-ready briefs.","https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Ftriage\u002FSKILL.md",[],[],{"skillNo":92,"name":93,"nameZh":6,"description":94,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":95,"categories":96,"keywords":97,"publishedAt":19},"01M77M469D7FDFHFHA2H65R1N7","to-tickets","Break a plan, spec, or the current conversation into a set of tracer-bullet tickets, each declaring its blocking edges, published to the configured tracker (edges as text in one file per ticket locally, or native blocking links on a real tracker).","https:\u002F\u002Fgithub.com\u002FRoeelfs\u002Fkeel\u002Fblob\u002Fmaster\u002F.claude\u002Fskills\u002Fto-tickets\u002FSKILL.md",[],[]]