Skip to content

Deep Research Prompt: OpenSpec Evaluation for Architecture Solution Design Workflow

Research Objective

I need a technically precise analysis of OpenSpec (by Fission AI) evaluated as a potential workflow framework for my architecture solution design process.

The question is: Should I adopt OpenSpec to improve my existing custom solution design workflow?

My Exact Setup

I have a Continuous Architecture Platform for a microservices-based system (19 services, 9 domains) with a custom solution design workflow:

  • Solution folder structure: architecture/solutions/_NTK-XXXXX-slug/ with subfolders for requirements, analysis, assumptions, capabilities, decisions (MADR format), guidance, per-service impacts, risks, and user stories
  • Workflow enforcement: Convention-based — documented in an ~800-line copilot-instructions.md file that the AI agent reads and interprets
  • AI tooling: GitHub Copilot Agent Mode (Claude Opus 4.6) — single tool, Copilot-specific instructions
  • Capability tracking: YAML-based capability changelog records L3 capability emergence per solution
  • Decision format: MADR (Markdown Any Decision Record) with required sections (Status, Date, Context, Decision Drivers, 2+ Options, Outcome, Consequences)
  • Portal: MkDocs Material site with generators that consume metadata YAML
  • CI/CD: GitHub Actions validating solution folder structure, YAML schema, and portal generation

Specific Questions to Research

1. OpenSpec's Workflow Model vs My Custom Workflow

OpenSpec structures work as "changes" containing artifacts (proposal, specs, design, tasks) with a dependency graph schema. My workflow uses a richer artifact set (requirements, analysis, assumptions, capabilities, MADR decisions, guidance, per-service impacts, risks, user stories).

  • Can OpenSpec's custom schema system accommodate my artifact types?
  • Can schemas enforce artifact-internal structure (e.g., MADR format within a decisions artifact)?
  • Can schemas handle variable-count subdirectories (e.g., impacts/impact.1/, impacts/impact.2/)?
  • Can the schema dependency graph trigger metadata rollups (e.g., updating capability-changelog.yaml on archive)?

2. Enterprise Evidence and Production Use

  • Who is actually using OpenSpec at scale? What is Comcast's specific use case?
  • What industries or organization types are adopting it?
  • Are there documented case studies beyond the marketing claims?
  • What does the community activity look like (GitHub issues, PR velocity, Discord/forum activity)?

3. Custom Schema Capabilities

  • What are the actual limits of OpenSpec's custom schemas?
  • Can I inject templates into generated artifacts (e.g., pre-populate MADR sections)?
  • How does the schema validation work — does it check artifact content or just file existence?
  • Can schemas define artifact-specific rules (e.g., "decisions.md MUST contain at minimum 2 options")?

4. AI Agent Integration Mechanics

  • What exactly does OpenSpec generate for GitHub Copilot integration?
  • Does it create copilot-instructions.md content, skill files, or something else?
  • Would OpenSpec's generated Copilot integration conflict with my existing copilot-instructions.md?
  • How do the slash commands (/opsx:propose, /opsx:apply, etc.) work in Copilot Agent Mode specifically?
  • Does the AI agent quality actually improve with OpenSpec's structured guidance vs. a well-written instruction file?

5. Risks and Maturity

  • OpenSpec is v1.2.0 from a startup (Fission AI). What is Fission AI's funding, team size, and sustainability trajectory?
  • Is there any indication of foundation governance (e.g., joining FINOS, Linux Foundation, Apache)?
  • What is the breaking change history between versions?
  • How does the MIT license interact with custom schema IP?

6. Minimum Viable Evaluation

  • What is the smallest possible PoC to test whether OpenSpec improves my workflow?
  • Can I run a single architecture scenario through OpenSpec and compare output quality?
  • What are the concrete setup steps for a GitHub Copilot + OpenSpec integration?

Output Format Requested

  1. Direct answers to each numbered question above with source citations
  2. A clear verdict on whether OpenSpec is worth evaluating for architecture solution design (not just feature development)
  3. Specific risks that would make adoption inadvisable
  4. If recommending evaluation: concrete next steps with expected effort

Important Constraints

  • Do NOT assume OpenSpec is designed for architecture work — verify whether anyone uses it for architecture governance vs. feature development
  • Do NOT rely on marketing claims — look for independent usage evidence
  • Do cite specific GitHub issues, commits, or documentation pages where possible