Design technical solution

Skill aryeko/agentic-workflow-kit/skills/design-technical-solution

Use before delivery-track planning when product or design context has meaningful technical complexity: new modules, data/query changes, AI prompts/tools/triggers, observability, migrations/deploy surfaces, security boundaries, or multi-system integration. Can start from a PRD, existing design docs, technical notes, or session context when enough scope exists. Writes a high-level technical solution document for plan-delivery-track to consume. Ingests rich context first, shows the flow, asks only blocking questions, and records safe assumptions.From its SKILL.md

Install

npx -y skills add aryeko/agentic-workflow-kit --skill design-technical-solution

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

5.6 KB, as published. Nobody here has run it

design-technical-solution

Author the technical solution gate between product intent and delivery-track planning. This skill defines the high-level technical "how" for complex product work before delivery stories are sliced into lightweight briefs. It can start from a PRD, existing design docs, technical notes, or session context when the supplied material is enough to identify scope, requirements, and technical boundaries.

References (read before acting)

Technical solution contract: ${CLAUDE_PLUGIN_ROOT}/references/technical-solution-contract.md
Technical solution template: ${CLAUDE_PLUGIN_ROOT}/references/templates/technical-solution-template.md
PRD contract: ${CLAUDE_PLUGIN_ROOT}/references/prd-contract.md
Story brief contract: ${CLAUDE_PLUGIN_ROOT}/references/story-brief-contract.md

Step 0 - Show the flow

Before doing the work, show the user the exact flow:

I will do: ingest PRD/context -> audit existing technical surfaces -> draft assumptions -> ask only blocking questions -> write technical solution -> self-review -> suggest /plan-delivery-track.

Then proceed unless a blocking decision is needed.

Step 1 - Resolve PRD and solution location

Read .workflow/config.yaml if present and take paths.prdsDir (default docs/prds).
Locate the PRD at <prdsDir>/<slug>/ when a PRD slug or path is supplied, and verify it conforms to prd-contract.md.
If no conforming PRD exists but the user supplied explicit design docs, technical notes, or session context with enough product scope and acceptance outcomes, derive a short kebab-case <slug> from the product/work name or ask the user to confirm one before writing. Continue only after the slug is known, and record the source material as assumptions. If neither a PRD nor sufficient context exists, stop and point the user at /define-product.
Write the technical solution at <prdsDir>/<slug>/technical-solution.md.
If that file already exists, switch to resume/extend mode. Never overwrite it without explicit confirmation.

Step 2 - Ingest context before asking

Read the PRD, existing design docs, technical notes, session context, relevant repo docs, existing solution/design docs, and source surfaces before interviewing. Summarize the material found and list the assumptions it supports.

Ask only blocking questions: questions whose answer would materially change module boundaries, data/query design, AI prompts/triggers/tools, migration/deploy surfaces, observability, security boundaries, or test strategy. For safe defaults, record safe assumptions in the document instead of interrupting the user.

Add an Assumptions and blockers pass: list safe assumptions first, then ask only questions that block a coherent high-level technical solution. Add an Artifact boundaries pass: confirm the technical solution owns high-level how, while tracker, story brief, detailed technical story spec, implementation plan, and runtime artifacts own downstream detail and evidence.

Step 3 - Draft the technical solution

Copy references/templates/technical-solution-template.md and fill every required section:

Context and existing surfaces
Technical requirements
System architecture diagram
Proposed modules/components
Data/query design
AI prompts/triggers/tools
Observability/events/metrics
Migration/deploy surfaces
Testing strategy
Open technical questions
Inputs for delivery tracker/story briefs

The technical solution must cite PRD acceptance-criteria IDs from 08-acceptance-criteria.md. It must include concrete inputs for plan-delivery-track: candidate story areas, sequencing constraints, file contention, validation expectations, and technical solution headings that story briefs should cite. When there is no PRD, cite equivalent product outcomes from the supplied external context and mark the PRD criteria column as context-derived until a PRD is created.

Step 4 - Self-review

Before writing, check:

Every ship-blocker PRD acceptance criterion has a technical requirement or an explicit non-technical reason it does not need one.
Migration/deploy, observability, testing, and data/query sections are present even when the answer is "not in scope."
Open questions are truly blocking. Move implementer-resolvable questions into story-brief inputs.
The document does not duplicate tracker status, owner, plan, PR fields, or detailed story-spec implementation decisions.

Step 5 - Write and hand off

Write <prdsDir>/<slug>/technical-solution.md, list the file written, and suggest /plan-delivery-track as the next step. plan-delivery-track consumes this technical solution plus the PRD; it owns tracker rows and lightweight story briefs.

Do not auto-commit.

Idempotency

Re-running is safe: detect an existing technical solution document, report missing or thin sections, and offer to fill only those sections. Do not clobber an existing document without confirmation.