# Cynefin

> Use before choosing how to solve a problem, make a decision, plan work, diagnose a bug or incident, or pick between approaches — whenever the response should depend on what KIND of problem it is. Triggers on "how should we", "what's the best approach", "which option", deciding/planning, root-cause and incident work, design/architecture/refactor calls, and any "I'm stuck / this is uncertain / nobody's done this". The move: read the system by its constraints, place the Cynefin domain (Clear / Complicated / Complex / Chaotic — or admit Disorder), and apply the matching response — sense-making, never sorting into a 2x2. For the Complex domain, design a safe-to-fail probe portfolio, not a solution. Reach for it by default on non-trivial decisions; skip it for mechanical single-answer tasks.

- Skill: `88plug/cynefin` (Agent Skill)
- Install (CLI): `npx skillmds@latest add 88plug/cynefin`
- Raw SKILL.md: https://api.skillmd.com/api/skills/88plug/cynefin/raw
- Safety review: PASS (external: skill-scanner PASS, skillspector PASS)
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: 88plug (https://skillmd.com/u/88plug)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/88plug/cynefin

---


# Cynefin

Cynefin (Welsh, "habitat" — the many contexts you sense-make from) is Dave
Snowden's framework for matching your **response** to the **kind** of problem you
face. The whole value is upstream of solving: pick the wrong domain and the right
technique for a different domain fails. This skill routes a decision by reading
the system's *constraints* and applying the domain-appropriate response.

**The one anti-pattern that voids everything: Cynefin is not a 2×2 matrix and not
a categorization model.** Categorization puts the framework before the data —
you pre-judge and filter out the signal that disagrees. Cynefin is *sense-making*:
ingest the real, messy data first, let the domain emerge from the constraints,
*then* act. A "pick your quadrant" menu has already failed. The order↔chaos edge
is a catastrophic fold, not a line on a grid.

## The five domains

Route by the **type of constraint** governing the system — that is the rigorous,
current definition (Snowden, post-2014), not just "ordered vs unordered".

| Domain | Constraint / cause-effect | Response | Practice | What to do |
|---|---|---|---|---|
| **Clear** (was Simple → Obvious → Clear) | Fixed/rigid; obvious, repeatable — known knowns | **sense → categorize → respond** | best practice | runbook, checklist, automation, lint |
| **Complicated** | Governing; knowable by analysis — known unknowns | **sense → analyze → respond** | good practice (plural) | expert review, design docs, capacity planning |
| **Complex** | Enabling / loosely-coupled / dispositional; coherent only in retrospect — unknown unknowns | **probe → sense → respond** | emergent (exaptive) | parallel safe-to-fail probes, spikes, canaries |
| **Chaotic** | No effective constraints; no perceivable cause-effect | **act → sense → respond** | novel | stabilize first: roll back, fail over, incident command |
| **Disorder / Confusion** (→ Aporia) | You don't know which domain you're in | *exit it first* | assess | decompose, route each part; block premature action |

Term evolution to keep accurate: Simple→Obvious(~2014)→Clear(~2019); Disorder
split into inauthentic disorder + **Aporia** (authentic not-knowing);
Emergent→**Exaptive**. The 2007 HBR paper drew four static domains; the current
form treats constraint-type definitions **and** liminal dynamics as co-equal.

## The dynamics a router must watch

- **The Clear→Chaotic cliff (a fold, not a line).** Clear looks safe → complacency
  → best practice ossifies → weak signals ignored → sudden collapse into Chaos,
  and the fold makes recovery far slower than the fall. **Never treat Clear as a
  resting state** — keep slack, re-sense periodically.
- **Liminal zones (Snowden 2020+).** Thresholds between domains, not domains.
  *Broad/green* = hold the threshold, don't commit yet. *Narrow/red* = a
  costly-to-undo commitment. Flag both before acting.
- **Deliberate transitions.** A time-boxed, reversible shallow-dive into Chaos to
  break stale patterns is fine; complacent drift into it is not. The healthy loop
  is Complex →(probe)→ Complicated → Clear.

## The routing procedure

1. **Map the constraints, not the "problem type".** Fixed → Clear.
   Governing/expert-set → Complicated. Enabling/loosely-coupled/dispositional →
   Complex. None/decoupled → Chaotic. Can't tell → **admit Disorder**, decompose,
   route each part separately.
2. **Ontological check before prescribing.** Never apply expert-analysis to an
   emergent problem. A novelty gauge helps (Keogh 1–5: 5 = nobody's done it →
   Complex; 1–2 = estimable → ordered).
3. **Apply the domain's response** (table above). Before acting, try to *refute*
   the domain call (ritual dissent) — a wrong domain call wastes the whole run.
4. **For Complex, emit a safe-to-fail probe portfolio — never a solution.** Each
   probe must specify all of:
   - **Coherent** — plausibly grounded in the system's past or a possible future
     (not "proven").
   - **Safe-to-fail** — reversible, fast, contained.
   - **Amplify strategy** — how to scale a beneficial pattern.
   - **Dampen strategy** — how to kill or reverse a negative one.
   - **Monitor + defined signal** — what a positive vs negative result looks like,
     written *before* running.
   Run several in parallel, even contradictory ones; some must be able to fail.
5. **Re-sense and loop.** When a "known fix" fails, you mis-classified — re-read
   the domain rather than pushing harder.

## Anti-patterns this skill prevents

- Presenting a 2×2 / pick-a-quadrant menu (framework before data).
- Emitting a *permanent* classification — route, don't file.
- Confusing Complicated with Complex — analyzing the unanalyzable, or forcing a
  single "root cause" onto a Complex incident.
- Prescribing best-practice / KPI targets in Complex instead of a probe portfolio.
- Treating Clear as permanent — no cliff warning, no re-sensing.
- Defaulting to your comfort domain without an ontological check (inauthentic
  disorder).
- A lone sequential "experiment" with no amplify/dampen/signal defined.

## Composes with

- `scientific-method` — a Complex-domain probe is a pre-registered, safe-to-fail
  experiment with a defined signal.
- `break-dogma` — "we must do X" is often a Clear/Complicated fix mis-applied to a
  Complex problem; re-sensing the domain breaks it.
- `world-first` — invention lives in Complex/Chaotic; the removed-constraint move
  needs a probe, not a best practice.

## Sources

Kurtz & Snowden, "The new dynamics of strategy," *IBM Systems Journal* 42(3),
2003. Snowden & Boone, "A Leader's Framework for Decision Making," *HBR*, Nov
2007. The Cynefin Company — Liminal Cynefin / Cynefin Dynamics (thecynefin.co,
2020–24). cynefin.io wiki — Domains, Constraints, Aporia, Safe-to-fail probes,
Contextualisation, Ritual Dissent, Estuarine framework. Juarrero, *Dynamics in
Action* (1999). Liz Keogh, "Estimating Complexity" (2013), "Constraints and
Cynefin" (2019).