| name | tech-design |
| description | Use when an approved spec needs a technical design -- architecture and technology choices grounded in compared options -- before planning. |
Tech Design
Design how to satisfy the approved Decision Spec: compare candidate approaches, choose one, and record the resulting design — architecture, components, and the technical decisions that bind it. Options research is the evidence; the chosen design is the product. Start from approved needs, not a favored tool.
This file is the phase contract — the rules the artifact must satisfy. Execution mechanics (how an executor runs and returns) live with the executor and the workflow engine, not here.
Execution
Delegated by default: the engine stays orchestrator and runs the tech-design-phase workflow — the autonomous loop of the designer authoring 02-TECH-DESIGN.mdx, the format gate, both reviewers judging independently in parallel, and capped rework — until it returns pass, needs-user, or the rework cap. The engine never authors or reviews in its own context; it only presents the result at the user gate. Under the same contract, fallbacks spawn the designer as a teammate or run in-situ.
Contract
- Input: the approved
01-DECISION-SPEC.mdx, plus any architecture answers the Spec interview recorded as design inputs.
- Output: one evolving root human artifact,
02-TECH-DESIGN.mdx.
- Research multiple option families: first-party implementation, capabilities of the existing runtime/platform, reusable in-repo skills/agents/components, and third-party packages.
- Third-party dependencies are references only until safety audit, compatibility check, wrapper design, and user approval.
- The chosen design follows from the scorecard: pick the winning option and commit it to a concrete shape — architecture sketch, component responsibilities, data flow. H3 subsections inside Chosen design are free-form.
- Key technical decisions records every binding choice beyond the headline option (libraries, protocols, storage, naming) as decision | choice | rationale rows; a decision without a stated alternative is a smell.
- Artifact UX assessment: how the design affects the user-facing artifact experience, under the work-item layout rule (root numbered human artifacts, underscore internals, no routine history).
- Findings that change the product need trigger a focused Spec addendum, never a silent edit to the spec.
Artifact shape and language
- Rich MDX, authored with
communicating-in-mdx loaded — reach for a component where it lands harder than prose, not for decoration. Contract-clean Markdown underneath: H1 title on line 1, H2 sections, no extra H2s, no frontmatter. Section names and order are normative in ../../contracts/tech-design.json.
- Structure over prose. Scorecard and Key technical decisions are always tables; other enumerable facts go in tables (option | tradeoff; need | coverage), parallel items in lists with bold lead-ins. Render architecture/data flow as a
```mermaid fence or <Mermaid> inside Chosen design; a <Callout> carries a load-bearing decision or a design-blocking risk. Prose only where narrative explains. A few sentences for simple sections, ~300 words for nuanced ones.
- Clean-slate, not a log. Write as a first draft by someone who knew the answer from the first prompt, not a record of how the discussion reached it. Per paragraph: would this exist in a one-shot draft? Cut what survives only because the conversation debated, corrected, or circled. Scrub the tells: replacing or superseding a "current state" on a from-scratch design; a dropped option used as a running comparison baseline (it lives only in Rejected alternatives); history words (settled, the old, obsolete, this pass, no longer, verified-not-speculation); the user wanted / rejected X — state the reason, not who decided.
- Each fact once; the Approval record is a pointer. No fact recurs across Scorecard, Key technical decisions, Rejected alternatives, and Approval record. The Approval record is the checkbox plus a one-line pointer to the decision trail (
_phases/) and any evidence — never a recap of the design, the most common log-residue failure. History and the cross-round <DocDiff> live on the gate page, not the body.
Format gate
Before the review gate: mdsmith check -c <plugin>/contracts/mdsmith.yml <work-item>/02-TECH-DESIGN.mdx, where <plugin> is the installed plugin root (the dir containing contracts/) and <work-item> is .workflow/<id>/. mdsmith reads MDX directly — JSX components and ```mermaid fences pass through. It enforces document shape (H1 on line 1, H2 sections in contract order, no frontmatter, no extra H2s): MDS020 is a contract violation, fix first; MDS023/MDS036/MDS056 are language budget, rework input; MDS025 (table alignment) is advisory and auto-fixable with mdsmith fix. The author runs the gate after each writing round and fixes structure before returning. Without mdsmith installed, verify sections manually against the contract JSON.
Scorecard
Derive the needs from the approved Decision Spec — its Constraints and Acceptance criteria sections, plus any need the spec's Goal makes explicit. One row per need, one column per option, each cell a short verifiable judgment. Do not reuse another work item's needs list; if the spec yields fewer than three needs, that is a Spec gap — raise it instead of padding.
Review gate
Run fixed reviewers. Their checklists are normative in this plugin's agents/<reviewer>.md; this table is only a map.
| Reviewer | Focus |
|---|
Reuse/Coverage Reviewer (agents/reuse-coverage-reviewer.md) | Candidate breadth, source coverage, needs mapping |
Fit/Risk Reviewer (agents/fit-risk-reviewer.md) | Design follows from evidence; capability fit, lock-in, safety/audit risk, reversibility |
The verdict enum, aggregation rule, and rework cap are machine truth in contracts/work-item.json, applied by phase-loop.js. The distinction reviewers judge: needs-user is a product question only the user can answer or a contradiction with stated intent; needs-rework is a fixable defect.
