| name | review-doc-orchestrator |
| description | Entry point for producing a branded review document. Classifies the request into one of five artifact types (user-story, architecture, PRD, risk-analysis, test-charter), gathers evidence scope, delegates content authoring to the matching artifact skill, and renders/validates the result. Use when someone asks for a reviewable, company-themed doc from analysis (e.g. "make an architecture review", "turn this into a risk register HTML"). |
review-doc-orchestrator
Routes a request to the right review-doc artifact skill and drives it through to
validated, rendered HTML. The model authors content (compact canonical
source); the renderer owns all chrome. Never hand-author HTML here.
Workflow
- Classify. Map the request to one artifact type (see
references/routing.md).
If the request is ambiguous, ask the user to confirm the type before authoring.
- Gather evidence scope. Collect links to code, ADRs, docs, incidents the
review should cite; record them as
evidence: [{label, href}].
- Pick theme + skin. Default theme (
assets/default-theme.json); the
renderer picks the skin from the artifact type (architecture/test-charter →
docs-site, PRD/user-story → editorial, risk-analysis → brief). Set
theme.layout only when the user wants a different look (e.g. an
executive brief of an architecture review).
- Delegate authoring. Invoke the matching artifact skill to produce compact
canonical YAML/JSON. Keep it content-only, 2–12 KB, one diagram per core section.
- Validate. Run
validate-branded-doc (schema first). Fix and re-author on failure.
- Render. Run
render-branded-html. Freeze diagrams / export PDF only if asked.
Routing (quick reference)
| If the request is about… | Artifact type | Skill |
|---|
| stories, acceptance criteria, scope gaps | user-story-review | user-story-review-doc |
| topology, interfaces, design trade-offs | architecture-review | architecture-review-doc |
| PRD coverage, goals, readiness | prd-review | prd-review-doc |
| risks, mitigations, residual risk | risk-analysis | risk-analysis-review-doc |
| exploratory testing missions, charters | test-charter | test-charter-review-doc |
Full routing heuristics and disambiguation rules: references/routing.md.
Artifact field/overlay detail: references/artifact-types.md.
Guardrails
- One artifact type per document. If a request spans two (e.g. architecture +
risk), produce two artifacts or confirm the primary one with the user.
- Output canonical source that passes schema validation before rendering.
- Default to interactive Mermaid; freeze only for archival/PDF. Diagrams are
expandable/zoomable in every skin — no extra work needed.
- Prefer widget section kinds over plain prose when data is structured
(callouts, stats, steps, cards, heatmap, meters — see the artifact skill's
Widgets section); overlay fields render automatically, never duplicate them.
- Keep this
SKILL.md thin — load references/* only when needed.
