Six use cases for v0.3.0 — Whyboard + Story templates + audit gate (live demos)

[seanhanca]

Show-and-tell thread for v0.3.0 templates and use cases. Reply with the chart you built; I'll start the thread with six live demos showing what the new Whyboard + Story templates and audit gates enable.

Each use case below is a one-click playground deep link. Click → spec loads → it compiles in the browser → you see byte-identical SVG to what @glyph/mcp@0.3.0 would return server-side. Edit the spec on the right pane to riff.

---

1. Diagnostic — "Why did Q3 revenue miss target?"

The classic agentic question: a metric dropped, the user wants to know why. A planner can pick this template as the starting shape, swap in real data via glyph_spec_patch, and emit a board the next reasoning hop can chain into. Anomaly · decompose · forecast — three branches the verb returns.

▸ Live demo: https://seanhanca.github.io/glyph/play/?example=wb-revenue-miss

{ "tool": "glyph_whyboard", "arguments": { "question": "Why did Q3 miss target?", "source_handle": "h_q3rev" } }

---

2. Diagnostic — "Why is API latency spiking?"

Same Whyboard shape, ops domain. /search and /feed jumped at 14:22 UTC; DB query time explains 78% of variance; recovery ETA ~22 min. The template makes the layout free — the agent only has to fill in branch summaries and link the leaf charts.

▸ Live demo: https://seanhanca.github.io/glyph/play/?example=wb-latency-spike

---

3. Narrative — Quarterly review storyboard

4-panel 2×2 grid: revenue trend → segment mix → cohort retention → growth forecast. Each panel a step badge + caption + chart slot. An LLM patches field-by-field via glyph_spec_patch; the layout never moves.

▸ Live demo: https://seanhanca.github.io/glyph/play/?example=story-quarterly-review

---

4. Narrative — Incident retrospective

Same 4-panel shape, postmortem domain: latency spike → error budget burn → root-cause attribution → recovery curve. Ready-made layout an SRE agent can populate the moment a P1 closes.

▸ Live demo: https://seanhanca.github.io/glyph/play/?example=story-incident-retro

---

5. Narrative — A/B test readout

Experiment-decision shape: traffic split → primary metric lift → secondary metrics → decision matrix. Drop in real exposure + outcome data and you've got the readout deck.

▸ Live demo: https://seanhanca.github.io/glyph/play/?example=story-ab-test-readout

---

6. Guardrails — audit gate on a refinement patch

This one is best shown in code, not in the playground. An agent has rendered a clean bar chart and now wants to "emphasize the dip" by tightening the y-axis:

{
  "tool": "glyph_spec_patch",
  "arguments": {
    "handle_id": "h_quarterly_revenue",
    "patches": [
      { "op": "replace", "path": "/layers/0/encoding/y",
        "value": { "field": "revenue", "scale": { "domain": [100, 200] } } }
    ]
  }
}

v0.2.0: silently rendered the misleading chart. v0.3.0:

{
  "error": "audit_regression",
  "message": "Patch would introduce HIGH-severity audit findings the original spec didn't have.",
  "regressions": [
    { "rule_id": "AUDIT-01", "severity": "high",
      "path": "/layers/0/encoding/y",
      "message": "Layer 0: bar chart y-axis domain starts at 100, not 0 — bar heights misrepresent magnitude." }
  ]
}

The agent has to either fix the patch or pass acknowledged: true to opt in to the regression. The iteration loop has a guardrail; misleading patterns no longer slip in silently.

---

Show me yours

Got a chart you built with 0.3.0? Reply with:

1. A one-line description of what it shows
2. A playground share URL (the Share button in the playground generates one)
3. Which audit rules you ran through (glyph_audit_spec output is welcome)

Templates are starting points, not constraints — happy to add the best community submissions to the manifest as new chips.

🚀 Read the 0.3.0 essay for the longer narrative on what the release enables.

🤖 All six templates and the audit gate were written, tested, and merged by Claude.