design-doc-reconciliation
Workflow for generating, reviewing, and reconciling design documents across multiple agents.
Skill by ejacklab
Install
Works with Claude Code, Claude apps, Codex, Cursor, Gemini CLI, GitHub Copilot, VS Code, OpenCode, Goose, Windsurf. Read the repository before you give any tool access to your data.
npx agentmarket install skill/ejacklab-design-doc-reconciliation
Asks which of your tools to add it to (Claude Code, Codex, Cursor, VS Code, Windsurf, Gemini CLI, Claude Desktop) and shows the change first. See https://agentmart.dev/cli.
Get the weekly diff
The weekly diff: one email a week with new agent tools, notable updates and grade changes, plus security alerts for tools you follow. Double opt-in, one-click unsubscribe.
About
Use cases: documents, design, data, productivity
Reviews
Write a reviewNo reviews yet. Installed it? Reviews come from verified installers only. How reviews work
Permissions
What this skill can do and what it asks for, from its SKILL.md. Checked Oct 25, 2026.
- Tools
- No tools found
- Secrets
- None asked for
- Install scripts
- None found
Network hosts it talks to 1
codeload.github.com· found in install command
Files it touches
~/.claude/skills/design-doc-reconciliation
What's inside
- Skill name
design-doc-reconciliation- Description
- Workflow for generating, reviewing, and reconciling design documents across multiple agents. Use when: (1) starting a new project with multiple design docs (OpenAPI, data models, data flows), (2) multiple agents need to produce design artifacts, (3) checking design docs for contradictions. NOT for: implementation (that's multi-agent-orchestration).
- License
- MIT
- Instructions
- 83 lines in SKILL.md · view on GitHub
Files (1)
- SKILL.md
SKILL.md (first lines)
# Design Doc Reconciliation ## The Problem When multiple agents write design docs independently, contradictions accumulate: - vtic Wave 1: **58 contradictions** across 6+ independently-written docs - Enums defined differently in OpenAPI vs data models - Field types mismatched between specs ## The Solution: Canonical Source First **Before any parallel generation, define the canonical source.** 1. **Choose canonical** — Usually OpenAPI spec (API-first) or data models 2. **Write canonical first** — One agent, one doc, reviewed 3. **Generate rest** — Parallel agents produce supporting docs 4. **Cross-review** — Separate agents check alignment to canonical 5. **Reconcile** — Fix contradictions BEFORE coding starts ## Workflow ### Phase 1: Canonical Source (Sequential) ``` 1. One agent writes the OpenAPI spec (or chosen canonical) 2. One agent reviews it 3. Commit as canonical ``` ### Phase 2: Parallel Generation ``` 1. Spawn 3-4 agents, each producing one doc: - Data models (align to OpenAPI) - Data flows (align to OpenAPI + data models) - Breakdown (align to all above) - Config schema (align to all above) 2. Each agent READS the canonical before writing ``` ### Phase 3: Cross-Review ``` 1. Spawn 2-3 review agents (GLM-5) 2. Each reviewer checks one doc against canonical 3. Output: numbered list of contradictions ``` ### Phase 4: Reconcile ``` 1. One agent reads all review reports 2. Fixes contradictions by aligning to canonical 3. The canonical doc is NEVER changed to match others ``` ### Phase 5: Lock ``` 1. Final review confirms zero contradictions 2. Commit all design docs 3. Design docs are now LOCKED — no changes without re-review ``` ## Reconciliation Rules - **Canonical is king** — All other docs align to it, never the reverse - **One field = one definition** — Same field must have same name, type, and constraints everywhere - **Zero tolerance for contradictions** — Even minor mismatches cause bugs downstream - **Review agents must READ the canonical** — Don't trust that generation agents got it right ## What Gets Reconciled | Doc | Aligns To | Common Issues | |-----|-----------|---------------| | OpenAPI spec | Self (canonical) | — | | Data models | OpenAPI | Field names, types, required/optional | | Data flows | OpenAPI + data models | Endpoint paths, request/response shapes | | Breakdown | All above | Module paths, function signatures | | Config schema | All above | Key names, defaults, types | ## Lesson: Cost of Skipping Reconciliation - vtic Wave 1 without reconciliation: 58 contradictions → 3 days of fix cycles - vtic Wave 1 with reconciliation: Zero contradictions → clean implementation - **Time saved by reconciling: >10x the time spent reconciling**
From ejacklab/vtic, MIT.
More from ejacklab/vtic
- 0.414score
self-harness
ejacklab
Design and use a workspace-local self-harness for multi-step work, planning, tracking, verification, and continuity.
SkillDFreshness: DClaude Code +9 - 0.409score
acp-management
ejacklab
Manage ACP (Agent Communication Protocol) sessions for Claude Code, Codex, and other coding harnesses.
SkillDFreshness: DClaude Code +9 - 0.409score
verify-before-trust
ejacklab
Verification protocol for agent outputs.
SkillDFreshness: DClaude Code +9 - 0.409score
multi-agent-orchestration
ejacklab
Orchestrate multiple coding agents (subagents and ACP harnesses) for parallel development.
SkillDFreshness: DClaude Code +9 - 0.408score
task-manager
ejacklab
Lightweight task tracking for cclow.
SkillDFreshness: DClaude Code +9 - 0.406score
planning
ejacklab
Workspace-local planning guide for cclow.
SkillDFreshness: DClaude Code +9 - 0.385score
find-skills
ejacklab
Helps users discover and install agent skills when they ask questions like "how do I do X", "find a skill for X", "is there a skill that can...", or express…
SkillDFreshness: DClaude Code +9 - 0.380score
test-hygiene
ejacklab
Guidelines for writing effective tests that actually catch bugs.
SkillDFreshness: DClaude Code +9 - 0.211score
scrum-master-lite
ejacklab
No description from the publisher yet.
SkillDFreshness: DClaude Code +9
Signals
- Stars
- 0
- Forks
- 0
- Open issues
- 0
- Last commit (this folder)
- Mar 21, 2026 (7 months ago)
From GitHub (fetched Oct 24, 2026).
Details
- License
- MIT
- Repository
- github.com/ejacklab/vtic
- First seen
- Oct 24, 2026
- Last checked
- Oct 24, 2026
- Source
- GitHubgithub-graphql
More from ejacklab
- 0.414score
self-harness
ejacklab
Design and use a workspace-local self-harness for multi-step work, planning, tracking, verification, and continuity.
SkillDFreshness: DClaude Code +9 - 0.409score
acp-management
ejacklab
Manage ACP (Agent Communication Protocol) sessions for Claude Code, Codex, and other coding harnesses.
SkillDFreshness: DClaude Code +9 - 0.409score
verify-before-trust
ejacklab
Verification protocol for agent outputs.
SkillDFreshness: DClaude Code +9 - 0.409score
multi-agent-orchestration
ejacklab
Orchestrate multiple coding agents (subagents and ACP harnesses) for parallel development.
SkillDFreshness: DClaude Code +9 - 0.408score
task-manager
ejacklab
Lightweight task tracking for cclow.
SkillDFreshness: DClaude Code +9 - 0.406score
planning
ejacklab
Workspace-local planning guide for cclow.
SkillDFreshness: DClaude Code +9
Cite this page
design-doc-reconciliation skill. Agent Mart. https://agentmart.dev/skills/ejacklab-design-doc-reconciliation (last updated 2026-10-24)