This repo collects lateral thinking techniques as agent skills. Contributions are welcome, and the bar is specific — please read it before writing.

• New techniques, with provenance. The technique must come from somewhere: a named author, a book, a documented practice. We do not accept invented "methods" dressed up with a citation-shaped sentence.
• New reference material for existing techniques: stimulus entries, analogy domains, SCAMPER questions, provocation templates.
• Worked examples from real sessions.
• Fixes: broken links, unclear steps, typos, structural problems.

Every technique must have honesty mechanics. Two things, non-negotiable:

1. An abandonment rule. A stated condition under which the technique gives up on a line of thought, and an instruction to show that abandonment to the user. If every stimulus, provocation, or flip in a session produces a usable idea, the session was faked — and a reader can tell. The visible failures are what make the successes credible.
2. A meta-pattern step. After the batch, scan across the results for the structural insight that recurs. The individual ideas matter less than what they have in common.

Worked examples come from real sessions. Run the technique. On a real problem. Include the abandonments that actually happened. Do not compose a plausible transcript — the difference is visible, and a fabricated example undermines the one thing this repo is for.

Descriptions use technique-specific triggers only. Generic stuck-ness language is reserved for the lateral router:

I'm stuck · going in circles · need fresh ideas · try a different angle

If a technique's description claims those phrases, it competes with the router for auto-invocation and routing becomes unpredictable. Describe what your technique is for, in its own words.

Every technique SKILL.md has exactly these sections, in this order:

---
name: <technique>
description: <technique-specific triggers only>
---

# <Technique Title>

## What this technique does
<3–4 sentences + provenance line: source, author, year.>

## Workflow
### Step 1: Confirm the target
<What a valid target looks like; one focused question to ask if unclear;
refusal rule: refuse requests to PERFORM analytical work (debug this,
review this code, implement this) — suggest an analytical approach
instead. Redesigning or ideating about such a process is a valid
creative target (e.g., "reinvent our code-review ritual" is in scope;
"review this PR" is not).>
### Step 2..N: <technique-specific>

## Honesty mechanics
<The technique's abandonment rule + meta-pattern step.>

## What NOT to do
<Bulleted anti-patterns.>

## References
<Relative links to files in ./references/.>

These skills run across different agents. Keep them portable:

• Frontmatter is name and description only. No other keys.
• No tool-specific syntax in any skill body: no agent names, no slash-commands, no vendor-specific formatting conventions. The single exception is the router's fallback, which may name the portable installer npx skills add danium/lateral-thinking.
• Relative paths only. Reference a sibling technique as ../<technique>/SKILL.md, never an absolute path.
• English only, plain markdown.

1. Open an issue first for a new technique. State the technique, its provenance, and which symptom it addresses that no existing technique covers. The lateral router's decision table maps one symptom to one technique — a new entry needs a symptom of its own.
2. Then open a PR following the template above.

For reference material and fixes, a PR without a prior issue is fine.