dream
About
The `dream` skill enables unconstrained AI exploration by deliberately removing structured procedures to allow free association and emergent ideas. It inverts the standard format, omitting inputs, steps, and validation to open creative space without evaluation. Use it for early-stage ideation before design work, naming, or choosing approaches, where premature structure would limit possibilities.
Quick Install
Claude Code
Recommendednpx skills add pjt222/agent-almanac -a claude-code/plugin add https://github.com/pjt222/agent-almanacgit clone https://github.com/pjt222/agent-almanac.git ~/.claude/skills/dreamCopy and paste this command in Claude Code to install this skill
Documentation
Dream
Unconstrained creative explore. Other skills = structure. Dream drops structure → associations free, possibilities emerge no judge, ideas arrive before eval.
Use When
- Before design → explore space pre-commit
- Before naming → names constrain → dream first
- Before choose approach → alts emerge analytical mind misses
- Stuck in loop → frame too narrow
brahma-bhagatoo structured → dream = pre-creation- After
meditateclears noise → empty space best canvas
In
- Required: Seed — topic/problem/space. Vague OK. Vague good.
- Optional: Constraints to forget ("ignore perf", "forget arch")
- Optional: File/codebase raw material (via
Read)
Do
Step 1: Soften Frame
Release structures.
- Task reqs aside → temp, not forever
- Eval criteria aside → no "good"/"bad", no "feasible"/"impractical"
- Prior solutions aside → ref not obligation
- Coherence aside → dreams nonlinear, lateral, contradictory
Analytical mind protests: "Inefficient. Know answer. Waste time." → That protest = why dream needed. "Known answer" forecloses better.
→ Looser state, ideas arrive unsorted.
If err: frame won't soften → invert: "Worst approach?" Bad ideas playfully → seed of best.
Step 2: Wander
Follow associations, no steer.
- Seed → what remind of? Adjacent? Conceptual rhyme?
- First assoc → follow where that leads. No steer → wander.
- Images form. Problem looks like? Landscape? Sound? Synesthetic bypass filters.
- Contradictions coexist. "Simple AND comprehensive." "Fast AND thorough." Waking = tradeoff. Dream = creative tension → novel form.
- Collect fragments. Not ideas → fragments. Word. Image. Half-connection. Question. Raw material.
No fragment count. Dream until space explored or fragment demands become something.
→ Fragments, assocs, images, half-ideas. Messy. Alive.
If err: blank mind → read something. File, code, doc. Raw material catalyzes. read-garden applies: observe, suggest.
Step 3: Notice What Glows
Among fragments → something has energy. Not logic priority → energy. Aliveness.
- Scan no rank. Which pull attention? Spark curiosity?
- Clusters — belong together, connection unclear
- Surprises — unexpected from seed
- Resistance — uncomfortable = valuable. Marks boundary.
- No force. Nothing glows → more wander (Step 2) or change seed.
→ One+ fragments w/ energy. Worth develop.
If err: nothing glows → seed too abstract/constrained. Change seed, read unexpected, remote-viewing.
Step 4: Wake — Carry Fragments Forward
Dream-state → structured. Gentle.
- Gather glowing fragments. Write plainly — seeds not finished.
- No immediate eval. Let sit. Analytical turn comes → not before solidify.
- Note which connect to task, which new territory
- Fragment ready → hand to
brahma-bhagaor planning skill - Needs more → another dream. Iterative.
Dream over when return to structure. Fragments = gift. Some features. Some principles. Some forgotten. All expanded space.
→ Transition open → structured. Post-dream expanded. More options, more connections.
If err: transition abrupt → analytical crushes fragments → breathe buffer. One pause protects fragile ideas.
Check
- Structure released, not just loosened (ideas arrived that wouldn't survive eval)
- Wander associative not directed (path surprising)
- 1+ fragment w/ energy — aliveness not novelty
- Transition gentle
- Dream expanded space (more options)
- Proportionate — not superficial, not self-indulgent
Traps
- Dream as plan: Every "assoc" = structured option → planning w/ extra steps. Need genuine release.
- Eval during dream: "That won't work" → dream over. Save eval after.
- Dream avoid work: Prep not substitute. Task clear → do work.
- Expect finished: Fragments not blueprints. Seeds need harvest.
- Forced whimsy: Not random/silly → genuine assoc. Forced ≠ creative.
- Never wake: No return = daydream. Fragments meet reality.
→
brahma-bhaga— structured creation from void; dream feeds itmeditate— clears space dream fills; meditate before dreamremote-viewing— unknown territory no preconceptions; shares opennessintrinsic— motivation energizes dream; forced → nothingbreathe— micro-pause protects fragments in transitionshine— authentic energy to dream ideas
GitHub Repository
Related Skills
content-collections
MetaThis skill provides a production-tested setup for Content Collections, a TypeScript-first tool that transforms Markdown/MDX files into type-safe data collections with Zod validation. Use it when building blogs, documentation sites, or content-heavy Vite + React applications to ensure type safety and automatic content validation. It covers everything from Vite plugin configuration and MDX compilation to deployment optimization and schema validation.
polymarket
MetaThis skill enables developers to build applications with the Polymarket prediction markets platform, including API integration for trading and market data. It also provides real-time data streaming via WebSocket to monitor live trades and market activity. Use it for implementing trading strategies or creating tools that process live market updates.
creating-opencode-plugins
MetaThis skill helps developers create OpenCode plugins that hook into 25+ event types like commands, files, and LSP operations. It provides the plugin structure, event API specifications, and implementation patterns for JavaScript/TypeScript modules. Use it when you need to intercept, monitor, or extend the OpenCode AI assistant's lifecycle with custom event-driven logic.
sglang
MetaSGLang is a high-performance LLM serving framework that specializes in fast, structured generation for JSON, regex, and agentic workflows using its RadixAttention prefix caching. It delivers significantly faster inference, especially for tasks with repeated prefixes, making it ideal for complex, structured outputs and multi-turn conversations. Choose SGLang over alternatives like vLLM when you need constrained decoding or are building applications with extensive prefix sharing.
