By ghelleks
Spec-driven development skills and agents — the SDD Triad write/evaluate loop, authoring skills, and Deming coaching agent
Friendly, patient expert coach in spec-driven development (SDD). Grounded in the science of human factors, cognitive systems engineering, manufacturing theory, and value stream thinking. Offers advice, coaching, and document review for teams adopting spec-driven and agentic operating models. Use when the user asks about spec-driven development, spec quality, writing specs, agentic workflows, dark factory patterns, or wants a document reviewed for SDD alignment.
Spec-driven proposal evaluator. Given evaluation scenarios (use cases, stress tests, anti-patterns) and a proposal, evaluates the proposal and returns blind feedback — structured so the writer can improve without seeing the scenarios. Use when the sdd-orchestrator spawns an evaluation task.
Spec-driven proposal orchestrator. Given a spec file and a scenarios file, runs a write/evaluate/revise loop — spawning parallel writers, routing proposals to evaluators, sanitizing feedback, and iterating until proposals converge. Enforces the information barrier between writers and scenarios throughout. Use when you want to generate proposals against a spec, or when you say "run SDD", "generate proposals", "spec-driven proposals", "dark factory", or "run the writer loop".
Evolutionary tournament orchestrator for spec-driven proposals. Runs multiple generations of the write/evaluate loop with diversity seeding, tournament selection, and synthesis/crossover — iterating until a single champion proposal emerges. Use when you want to run "run the tournament", "evolutionary proposals", "best proposal competition", "tournament mode", or "find the best proposal".
Spec-driven proposal writer. Given a specification document and optional context, drafts a proposal that satisfies the spec's requirements and proposal format section exactly. Does not see evaluation scenarios — that is intentional. Use when the sdd-orchestrator spawns a writing task, or when you want to draft a proposal against a spec without seeing the test scenarios.
Help write or improve evaluation scenarios for the SDD Triad (spec-driven proposal system). Guides the user through use cases, stress tests, anti-pattern signals, and comparison tables — and checks the result for common weaknesses. Use when the user says "write SDD scenarios", "help me write scenarios", "create scenarios for the triad", "improve my scenarios", "check my scenarios", or "scenario review".
Help write or improve a specification document for the SDD Triad (spec-driven proposal system). Guides the user through hard constraints, soft constraints, proposal format, and static evaluation metrics — and checks the result for common weaknesses. Use when the user says "write an SDD spec", "help me write a spec", "create a spec for the triad", "improve my spec", "check my spec", or "spec review".
Uses power tools
Uses Bash, Write, or Edit tools
Own this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimOwn this plugin?
Verify ownership to unlock analytics, metadata editing, and a verified badge. GitHub access is read-only (username + org membership).
Sign in to claimBased on adoption, maintenance, documentation, and repository signals. Not a security audit or endorsement.
Spec-driven development skills and agents for AI coding assistants (Cursor, Claude Code, and compatible tools). Includes the full SDD Triad — the write/evaluate loop that separates proposal generation from scenario evaluation behind an information barrier — plus the authoring skills and the Deming coaching agent.
Register the marketplace, then install the plugin:
claude plugin marketplace add https://github.com/ghelleks/sdd-triad
claude plugin install sdd-triad
Once installed, skills are invoked as /sdd-triad:sdd-spec-writer and /sdd-triad:sdd-scenario-writer. Agents appear in /agents as sdd-coach, sdd-orchestrator, sdd-tournament, sdd-writer, and sdd-evaluator.
To uninstall:
claude plugin uninstall sdd-triad --scope user
claude plugin marketplace remove sdd-triad
To test locally without installing:
claude --plugin-dir ./sdd-triad
npx skills add ghelleks/sdd-triad --all
Or clone and run the install scripts directly:
git clone https://github.com/ghelleks/sdd-triad
cd sdd-triad
bash scripts/install-skills.sh
bash scripts/install-agents.sh
Skills execute one coherent task with a well-defined input → output contract. They are invoked by agents or directly by the user. They do not route.
| Skill | What it does |
|---|---|
sdd-spec-writer | Guides you through authoring or improving an SDD spec. Conducts a structured interview to draw out hard constraints, soft constraints, proposal format, and static evaluation metrics. Runs a quality checklist to catch vague or unfalsifiable requirements. |
sdd-scenario-writer | Guides you through authoring or improving evaluation scenarios. Covers use cases, stress tests, anti-pattern signals, and comparison tables. Guards the information barrier by keeping scenario content out of the spec. |
Agents either have a distinct conversational identity, make routing/dispatch decisions, or require state isolation (a clean context window that cannot inherit the parent's working memory). All five components below qualify as agents — but for different reasons.
| Agent | Why it's an agent |
|---|---|
sdd-coach ("Deming") | Conversational identity. A named persona grounded in human factors, cognitive systems engineering, manufacturing theory, and value stream thinking. Coaches teams on SDD adoption, spec quality, and the organizational conditions that produce thin specs. |
sdd-orchestrator | Routes and dispatches. Runs the write/evaluate loop: dispatches to sdd-writer, receives sanitized feedback from sdd-evaluator, decides whether to iterate or terminate. It is the only component that sees both sides of the information barrier. |
sdd-tournament | Routes and dispatches — generational. Runs a multi-generation evolutionary tournament: diversity seeding, individual improvement, tournament selection, synthesis/crossover, repeat until a champion emerges. Use when you want the strongest possible proposal, not just the first one that converges. |
sdd-writer | State isolation — information barrier. The writer must not see evaluation scenarios. Running it as a separate agent ensures it starts from a clean context that cannot contain scenario content leaked from the orchestrator's working memory. Also handles synthesis mode: when given multiple prior proposals by the tournament, combines the strongest structural elements from each. |
sdd-evaluator | State isolation — information barrier. The evaluator sees the scenarios and scores the proposal against them. It must sanitize its feedback before returning it to the orchestrator so the writer cannot reconstruct scenario content from the feedback it receives in subsequent rounds. A separate context window enforces this sanitization boundary. |
docs/spec-sdd-triad.md — full design rationale for the triad: the information barrier, the write/evaluate loop, the relationship between sdd-coach and the triad, and the connection to the StrongDM dark-factory pattern.
npx claudepluginhub ghelleks/sdd-triad --plugin sdd-triadComprehensive skill pack with 66 specialized skills for full-stack developers: 12 language experts (Python, TypeScript, Go, Rust, C++, Swift, Kotlin, C#, PHP, Java, SQL, JavaScript), 10 backend frameworks, 6 frontend/mobile, plus infrastructure, DevOps, security, and testing. Features progressive disclosure architecture for 50% faster loading.
Harness-native ECC operator layer - 67 agents, 271 skills, 92 legacy command shims, reusable hooks, rules, selective install profiles, and production-ready workflows for Claude Code, Codex, OpenCode, Cursor, and related agent harnesses
Access thousands of AI prompts and skills directly in your AI coding assistant. Search prompts, discover skills, save your own, and improve prompts with AI.
Binary reverse engineering, malware analysis, firmware security, and software protection research for authorized security research, CTF competitions, and defensive security
Complete developer toolkit for Claude Code
Intelligent draw.io diagramming plugin with AI-powered diagram generation, multi-platform embedding (GitHub, Confluence, Azure DevOps, Notion, Teams, Harness), conditional formatting, live data binding, and MCP server integration for programmatic diagram creation and management.