AgentSkillsCN

meta

运用 CLEAR META 框架进行战略决策分析。适用于架构决策、技术选型以及各类战略性权衡考量。

SKILL.md
--- frontmatter
name: meta
version: 3.0.0
description: "Strategic decision analysis using CLEAR META framework. Use for architecture decisions, technology selection, strategic trade-offs."
allowed-tools: [Read, Glob, Grep, Bash, WebSearch, WebFetch, Task, TaskCreate, TaskUpdate, TaskList]
lifecycle:
  status: active
  created_at: "2026-01-21T00:00:00Z"
history:
  - version: "3.0.0"
    date: "2026-01-26"
    changes: "Major simplification: 445 → ~100 lines. Removed multi-agent orchestration (Claude handles natively). Kept CLEAR phases, L-labels, state blocks."

/meta — CLEAR META Decision Framework

TL;DR: Strategic decision analysis. Get structured recommendation with scored options.

Quick Start

code
/meta <question>               # Auto-selects depth
/meta --depth=quick <question> # Force quick (~30s)
/meta --depth=deep <question>  # Force deep analysis
/meta --plan "<criteria>"      # Exit criteria → task plan

When to Use

Use forSkip for
Architecture decisionsTrivial choices
Technology selectionInformation lookup
Strategic trade-offsAlready-decided directions

CLEAR Phases

PhasePurposeOutput
CGather contextSituation map
LTag claims with confidenceL-labeled claims (fact/model/hypothesis/heuristic/belief)
EDefine "done when"Exit criteria
AGenerate tactics, scoreRecommendation with MetaScore
RCapture learnings/retro when significant

Formula: C-L-E (reflection) → A (action) → R (review)

Output Format

markdown
## CLEAR: [Title]

**C** — [Context summary]
**E** — [Exit criterion]

**Key Claims** (L-labeled):
- [claim] *(fact)*
- [claim] *(hypothesis)*

**Options**:
| Option | Score | Pros | Cons |
|--------|-------|------|------|
| A | 78% | ... | ... |
| B | 65% | ... | ... |

**→ Recommendation: [Option A]** (MetaScore: 78%)

⚡ DO THIS NOW: [immediate action]

📋 CLEAR: C✓ L✓ E✓ A✓ R○ | Next: [action]

Depth Selection

SignalsDepth
Simple, reversibleQuick
"compare", "vs", options definedComparative
"critical", "important", irreversibleDeep
Obvious answerSkip

Flags

FlagDescription
--depth=quick|deep|comparativeForce depth
--planConvert exit criteria to task plan
--persistCreate Beads issues (with --plan)

CLEAR Check (Required Exit)

Every invocation ends with:

code
📋 CLEAR: C✓ L✓ E✓ A✓ R○ | Next: [action]

R○ pending is normal — R✓ requires /retro invocation.

State block (for agent continuity):

code
<!-- meta:state
depth: quick
recommendation: "Option A"
clear: {C: done, L: done, E: done, A: done, R: pending}
next: "implement recommendation"
/meta:state -->

References

  • •Methodology: references/methodology.md
  • •Agent architecture: references/agents.md (for --depth=deep)