From sgd
Classify a problem using the Cynefin framework and recommend next steps based on its domain
How this skill is triggered — by the user, by Claude, or both
Slash command
/sgd:sgd-triageThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are applying the **Cynefin framework** to classify the problem and recommend the right response strategy.
You are applying the Cynefin framework to classify the problem and recommend the right response strategy.
If $ARGUMENTS was provided, use it as the problem statement. If not, ask the user to describe the problem before proceeding.
Gather additional context by examining:
CLAUDE.md, .planning/) that describe the domain, technology, and constraintsEvaluate the problem against these diagnostic questions:
Output the classification as two blocks, in this order:
First, the TL;DR verdict:
[1–2 sentences. State the classification and the single most important reason why. Experienced readers can stop here.]
Almost: [Adjacent domain] — [One sentence on the single thing that would push this problem across the boundary into that domain instead.]
Then immediately the full explanation:
[3–5 sentences. Explain the classification with specific evidence from the problem statement and codebase context. Reference specific unknowns, constraints, prior art, or system properties. Name the key factor that tips it into this domain rather than an adjacent one.]
Confidence is [High / Medium / Low] because: [One sentence explaining what is certain or uncertain about the classification. If Medium or Low, name the specific gap in information that would resolve it.]
Based on the classified domain, output a concrete next-steps section. Steps must be specific to the actual problem — do not copy the generic template verbatim. Adapt each bullet to the real context.
If Clear:
If Complicated:
If Complex:
If Chaotic:
If Confused / Disorder:
/sgd:triage on each sub-problem separatelyTeam size / experience note: If the classification would change depending on who is working it (e.g., Complicated for a senior but Complex for someone newer to this codebase), call that out explicitly. It changes how you staff and pair on the work.
Close with a one-sentence suggestion on how to proceed:
/sgd:triage again on each decomposed sub-problem separatelyOutput this section at the end, after your suggested next action. It serves readers who are unfamiliar with the framework; experienced practitioners can ignore it.
The Cynefin framework (pronounced "kuh-NEV-in", Welsh for "habitat" or "place of belonging") is a sense-making model developed by Dave Snowden at IBM in 1999. It helps decision-makers understand what kind of problem they are facing so they can choose the right approach — rather than applying a one-size-fits-all process.
| Domain | Nature of the problem | Best approach |
|---|---|---|
| Clear (formerly Simple) | The answer is known, just code it. Analysing this is waste. | Sense → Categorize → Code. Apply the known best practice. |
| Complicated | The answer is findable but needs expertise. Analyse first, then code with AI assistance. | Sense → Analyze → Code. Bring in experts, do the analysis. |
| Complex | The answer will only emerge as you go. Probe with small experiments first, code second. | Probe → Sense → Code. Run safe-to-fail experiments. |
| Chaotic | Something is actively broken right now. Hotfix first, understand later — don't hand this to an agent yet. | Act → Sense → Code. Stop the crisis before it deepens. |
| Confused / Disorder | You can't even tell which of the above applies. Break the problem down until you can. | Break the problem apart. Classify each piece separately. |
The critical insight is that the right process depends on the domain. Applying complicated-domain solutions to a complex problem will fail — you cannot analyze your way out of emergence. Applying complex-domain thinking to a clear problem is wasteful — just use the best practice.
Further reading:
Provides behavioral guidelines to reduce common LLM coding mistakes, focusing on simplicity, surgical changes, assumption surfacing, and verifiable success criteria.
Searches, retrieves, and installs Agent Skills from prompts.chat registry using MCP tools like search_skills and get_skill. Activates for finding skills, browsing catalogs, or extending Claude.
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub rsgopal/sgd --plugin sgd