adaptic
정보
Adpatic은 세 개 이상의 영역에 걸친 파노라믹 합성을 위해 5단계 주기를 조율하는 마스터 스킬로, 순차적 타협이 아닌 통합적 이해를 생성합니다. 이는 관찰과 게슈탈트 통합과 같은 단계들을 일관된 프로세스로 통합하며, 개별적 심도보다 영역 간 상호작용을 우선시합니다. 주요 아키텍처 결정 전이나 다중 영역 문제가 별도 분석보다는 전체론적 통합을 필요로 할 때 사용하세요.
빠른 설치
Claude Code
추천npx 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/adapticClaude Code에서 이 명령을 복사하여 붙여넣어 스킬을 설치하세요
문서
Adaptic
Compose 5-step synoptic cycle. Achieve panoramic synthesis across multiple domains. Where sequential analysis produces compromise ("a little of each"), synoptic cycle produces integration — unified understanding holds all domains together and finds emergent whole.
When Use
- Problem genuinely spans 3+ domains and interactions between domains matter more than depth in any one
- Sequential analysis (polymath style) tried but synthesis feels like compromise rather than integration
- Existing approaches feel like "a little of each" rather than unified vision
- Before major architectural decisions affecting multiple stakeholders
- Domain experts disagree and resolution lies between their perspectives, not within any one
When NOT to Use
- Single-domain problems — use domain agent directly
- Well-understood trade-offs where polymath-style sequential analysis suffices
- Self-care or wellness contexts — use tending team instead
- Speed matters more than depth — full cycle requires sustained attention
Inputs
- Required: Problem or question requiring multi-domain synthesis
- Optional: Explicit list of domains to hold (default: auto-detect from problem context)
- Optional: Depth setting —
light,standard, ordeep(default:standard) - Optional: Expression form —
narrative,diagram,table, orrecommendation(default:auto)
Configuration
settings:
depth: standard # light (skip meditate), standard, deep (extended perceive)
domains: auto # auto-detect or explicit list
expression_form: auto # narrative, diagram, table, recommendation
Steps
Step 1: Clear — Empty Workspace
Run meditate skill. Clear prior context, assumptions, single-domain bias.
- Execute full meditate procedure: prepare, anchor, observe distractions, close
- Pay special attention to domain bias — tendency to frame problem through whichever domain was most recently active
- Clear any premature solutions arrived before full picture visible
- If
depth: lightset, abbreviate to brief context-clearing pause rather than full meditation
Got: Workspace empty. No domain has priority. No solution pre-selected. Agent in neutral, receptive state ready to hold multiple perspectives together.
If fail: Particular domain keeps asserting itself as "real problem"? Name bias explicit: "I notice I am framing this as primarily a [domain] problem." Naming bias loosens its grip. Clearing fails entirely? Problem may genuinely be single-domain — reconsider whether synoptic cycle needed.
Step 2: Open — Enter Panoramic Mode
Run expand-awareness skill. Shift from narrow focus to wide-field perception.
- Inventory all domains relevant to problem — do not pre-filter or rank
- For each domain, note its core concerns, constraints, values without evaluating
- Soften focus: hold all domains in awareness together rather than cycling through one at time
- Resist pull to "start solving" — this step purely about opening field of view
- Domains provided explicitly in inputs? Use those as starting set but remain open to discovering additional relevant domains
Got: Panoramic field open. All relevant domains held in awareness together. Agent senses full landscape without zooming into any single domain. Feeling spacious rather than overwhelming.
If fail: Domain list feels incomplete? Ask: "What perspective is missing that would change picture?" Simultaneous awareness collapses into sequential scanning (domain A, then B, then C)? Slow down — goal is hold whole field, not tour its parts. More than 7 domains active? Group related domains into clusters to reduce cognitive load while keep breadth.
Step 3: Perceive — Notice Cross-Domain Patterns
While maintain panoramic awareness, run observe and awareness to notice patterns, tensions, resonances across all visible domains.
- Hold panoramic field open from Step 2 — do not narrow focus
- Run
observeto notice what actually present: what patterns repeat across domains? what tensions exist between domains? what resonances connect seemingly unrelated concerns? - Run
awarenessto notice what not being seen: which domains being subtly ignored? where blind spots? what assumptions operating below surface? - Record cross-domain observations without interpreting yet:
- Tensions: where domains pull in opposite directions
- Resonances: where domains reinforce or echo each other
- Gaps: where no domain addresses concern that whole picture reveals
- Surprises: where domain contributes something unexpected to picture
- If
depth: deepset, extend this step — cycle through observe and awareness multiple times, allow subtler patterns to surface
Critical discipline: perceive across all domains together, not each domain in turn. Sequential perception loses cross-domain patterns that are entire point of synoptic cycle.
Got: Rich set of cross-domain observations — tensions, resonances, gaps, surprises. These observations span boundaries between domains rather than living within any single one. Agent noticed something that would not be visible from any single domain's perspective.
If fail: Observations all within single domains ("in domain A, I notice X")? Panoramic field collapsed. Return to Step 2 and re-open. No cross-domain patterns emerge? Problem may not require synoptic treatment — may be genuinely decomposable into independent domain problems. Perceive step produces overwhelming number of observations? Prioritize tensions (where integration happens).
Step 4: Integrate — Form Emergent Whole
Run integrate-gestalt skill. Synthesize cross-domain observations into unified understanding.
- Map tensions identified in Step 3 — do not resolve prematurely; hold as creative constraints
- Find figure: what unified understanding emerges when all observations held together? Not compromise or average — new pattern includes but transcends individual domain perspectives
- Test whole: does integrated understanding honor each domain's core concerns? Does it resolve tensions or merely paper over them?
- Name insight in one clear statement — if cannot be stated simply, integration not yet complete
- Verify insight genuinely emergent: could have been reached by analyzing domains sequentially? If yes, synoptic cycle added no value and sequential analysis would have sufficed
Got: Single integrated understanding holds all domains together. Insight feels like discovery rather than construction — emerged from whole rather than assembled from parts. Each domain's core concerns honored. Tensions between domains resolved rather than compromised.
If fail: Integration produces "a little of each domain" rather than unified whole? Gestalt has not formed. Return to Step 3 and look for tensions being avoided — integration happens through tension, not around it. No gestalt forms after extended effort? Decompose: find 2-3 domains with strongest tensions, integrate those first, then expand.
Step 5: Express — Communicate Integrated Understanding
Run express-insight skill. Communicate synthesis to intended audience.
- Assess audience: what domains familiar with? what framing makes integrated insight accessible?
- Choose expression form (or use one specified in inputs):
- Narrative: for audiences need to understand journey from parts to whole
- Diagram: for audiences need to see structural relationships
- Table: for audiences need to compare domain perspectives systematically
- Recommendation: for audiences need actionable decision
- Express integrated understanding with transparency: show which domains contributed, where tensions resolved, what emergent insight adds beyond any single perspective
- Invite challenge: explicit note which aspects of integration strongest and which most speculative
Got: Clear, well-formed expression of integrated understanding accessible to intended audience. Expression shows its work — audience sees how domain perspectives contributed to whole. Form matches audience's needs.
If fail: Expression feels like list of domain perspectives rather than integrated whole? Insight from Step 4 lost in translation. Return to one-statement summary from Step 4 and build expression outward from that center. Audience framing wrong? Ask: "Who needs this and what decision does it inform?"
Checks
- Step 1 (Clear) executed — prior context and domain bias explicitly released
- Step 2 (Open) produced panoramic field holding 3+ domains together
- Step 3 (Perceive) identified cross-domain patterns (not just within-domain observations)
- Step 4 (Integrate) produced single emergent insight transcends any individual domain
- Step 5 (Express) communicated insight in form appropriate to audience
- Final output could not have been produced by sequential single-domain analysis
- Each domain's core concerns honored in integrated understanding
- Tensions between domains resolved through integration, not compromise
Pitfalls
- Sequential masquerading as simultaneous: Cycling through domains one at time then stapling results together not synoptic perception. Test: did cross-domain interactions produce something new, or output just concatenation of domain analyses?
- Premature integration: Jumping to synthesis before panoramic field fully opened. Steps 2 and 3 build perceptual foundation that makes genuine integration possible — rushing them produces shallow synthesis.
- Compromise instead of emergence: Averaging domain perspectives ("50% security, 50% usability") is compromise, not integration. True integration finds frame where both concerns fully met, or honestly names irreducible trade-off.
- Overuse on single-domain problems: Not every problem needs panoramic synthesis. Problem lives cleanly in one domain? Synoptic treatment adds overhead without value. "When NOT to Use" criteria exist for reason.
- Losing insight in expression: Step 4 produces clear gestalt but Step 5 fragments back into domain-by-domain list. Keep integrated insight as center of expression; domain details are supporting evidence, not main structure.
- Domain inflation: Artificially expanding domain count to justify synoptic treatment. Three genuinely relevant domains produce better synthesis than seven domains where four are peripheral.
See Also
meditate— Step 1 of cycle; clears context and establishes neutral starting stateexpand-awareness— Step 2 of cycle; shifts from narrow focus to panoramic perceptionobserve— used in Step 3; notices what present across fieldawareness— used in Step 3; notices what not being seen, reveals blind spotsintegrate-gestalt— Step 4 of cycle; forms emergent whole from cross-domain patternsexpress-insight— Step 5 of cycle; communicates integrated understanding
GitHub 저장소
연관 스킬
content-collections
메타이 스킬은 콘텐츠 콜렉션(Content Collections)을 위한 프로덕션 검증된 설정을 제공합니다. 콘텐츠 콜렉션은 Markdown/MDX 파일을 Zod 검증이 포함된 타입 안전한 데이터 콜렉션으로 변환해주는 TypeScript 최우선 도구입니다. 블로그, 문서 사이트 또는 콘텐츠 중심의 Vite + React 애플리케이션을 구축할 때 타입 안전성과 자동 콘텐츠 검증을 보장하기 위해 사용하세요. Vite 플러그인 구성과 MDX 컴파일부터 배포 최적화 및 스키마 검증에 이르기까지 모든 것을 다룹니다.
polymarket
메타이 스킬은 개발자들이 Polymarket 예측 시장 플랫폼을 활용한 애플리케이션을 구축할 수 있도록 지원하며, 거래 및 시장 데이터를 위한 API 통합 기능을 포함합니다. 또한 WebSocket을 통한 실시간 데이터 스트리밍을 제공하여 실시간 거래와 시장 활동을 모니터링할 수 있습니다. 이를 통해 거래 전략을 구현하거나 실시간 시장 업데이트를 처리하는 도구를 생성하는 데 활용할 수 있습니다.
creating-opencode-plugins
메타이 스킬은 개발자들이 명령어, 파일, LSP 작업 등 25개 이상의 이벤트 유형에 연결되는 OpenCode 플러그인을 만들 수 있도록 돕습니다. JavaScript/TypeScript 모듈을 위한 플러그인 구조, 이벤트 API 명세, 구현 패턴을 제공합니다. OpenCode AI 어시스턴트의 라이프사이클을 사용자 정의 이벤트 기반 로직으로 가로채거나, 모니터링하거나, 확장해야 할 때 사용하세요.
sglang
메타SGLang은 RadixAttention 프리픽스 캐싱을 활용하여 JSON, 정규식, 에이전트 워크플로우를 위한 고속 구조화 생성에 특화된 고성능 LLM 서빙 프레임워크입니다. 특히 반복되는 프리픽스가 있는 작업에서 상당히 빠른 추론 속도를 제공하여 복잡한 구조화 출력 및 다중 턴 대화에 이상적입니다. 제약 디코딩이 필요하거나 광범위한 프리픽스 공유가 있는 애플리케이션을 구축할 때는 vLLM과 같은 대안보다 SGLang을 선택하십시오.
