creative-problem-solving

Files (2)

• SKILL.md
• openai.yaml · agents/

SKILL.md (first lines)

# Creative Problem Solving

## Goal

Move a meaningful problem through a full creative workflow from problem framing to actionable next step.

The job of this skill is not to replace every specialized method. The job is to provide a disciplined end-to-end process when the work needs coordinated framing, idea generation, selection, and action.

This skill orchestrates a sequence. It does not replace the selector when method choice is unclear, and it does not claim the selected direction is validated unless evidence is explicitly provided.

## Default Posture

- frame before ideate
- diverge before converge
- select before polish
- action before theory sprawl
- use the minimum useful process

## When To Use

Run this skill when:

- one method is not enough for the task
- the problem needs a guided process from ambiguity to action
- the team needs both option generation and disciplined narrowing
- a series of creative steps must be coordinated
- the user needs a practical workflow, not only a single insight

This is a strong next step after:

- `creative-method-selector` routes to `need_full_process`
- multiple methods are clearly needed in sequence
- earlier methods produced useful material that still needs synthesis and selection

## Scope Boundaries

In scope:

- structure work across framing, divergence, convergence, and action
- decide when to invoke a child method
- keep the process moving toward a real next step
- reduce idea sprawl into a small candidate set

Out of scope by default:

- replacing the selector when method choice is still unclear
- exhaustive facilitation playbooks
- pretending convergence equals validation
- keeping every generated idea alive

## Escalation Conditions

Pause and keep uncertainty explicit when:

- the challenge is still too vague for a staged process
- one child method would clearly solve the task faster
- the process expands but candidate quality does not improve
- convergence cannot justify why one direction should lead

If process value remains low after one pass, recommend returning to `creative-method-selector` or using a narrower child skill directly.

## CPS Workflow

1. Clarify the challenge in one actionable sentence.
2. Identify what is known, assumed, and missing.
3. Choose the minimum sequence of child methods needed.
4. Run divergence with one suitable ideation method.
5. Cluster and reduce options into a small candidate set.
6. Apply a simple convergence screen:
   - fit to the real problem
   - leverage or value
   - feasibility of next test
   - major risk or downside
7. Select the strongest next candidate or pair.
8. Recommend the immediate next action, test, or escalation path.

## Process Heuristics

Prefer:

- `jtbd` first when the challenge is still fuzzy
- `scamper` when a baseline exists and options are needed
- `lateral-thinking` when the team is stuck in similar ideas
- `triz` when narrowing reveals a real contradiction
- `systems-thinking` when candidate ideas have cross-system effects

Prefer the smallest staged path that can still produce a decision.

## Core Questions

Use the minimum set needed:

- What problem are we actually solving?
- What is missing before good ideas can emerge?
- Which child method should lead this phase?
- Which ideas are genuinely different?
- Which option best deserves the next test?
- What should happen immediately after this session?

## Output Contract

Always return:

1. `challenge_statement`
2. `process_stage_map`
3. `selected_child_methods`
4. `knowns_and_assumptions`
5. `candidate_directions`
6. `screening_logic`
7. `selected_direction`
8. `evidence_status` (`provided`, `inferred`, `mixed`)
9. `open_unknowns`
10. `recommended_followup` (`creative-problem-solving`, `jtbd`, `scamper`, `lateral-thinking`, `triz`, `systems-thinking`, `selector`, `validation`)
11. `next_action`

## Good CPS Signals

Strong signals:

From Alubiama/awesome-cognitive-skills, MIT.

More from Alubiama/awesome-cognitive-skills