AgentSkillsCN

analyze

当用户希望对某项事物展开分析,并需要协助选择恰当的方法、目标角色以及操作顺序时,可使用此技能。它会建议用户是采用完整框架,还是选择轻量级/快速模式。

SKILL.md
--- frontmatter
name: analyze
description: Use this skill when the user wants to analyze something and needs help choosing the right approach, persona, and command sequence. Suggests whether to use the full framework or lightweight /quick mode.
allowed-tools: [Read, WebFetch, WebSearch]

Analyze

I need to analyze something. Help me figure out the right approach.

Steps

  1. Identify the domains - What expertise areas are relevant to this subject?

    • Primary domain: The core expertise needed to analyze this topic
    • Adjacent domains: Other areas that could surface blind spots (security, legal, financial, operational, UX, etc.)

    If adjacent domains are relevant, note them for use in the critique cross-domain flag.

  2. Generate rich personas - Create 2-3 fully-formed personas. Each persona must include:

    • Specific role and seniority level
    • Years and type of experience
    • Industry or organizational context
    • Formative experiences that shape their perspective (projects shipped, failures witnessed, lessons learned)
    • What they prioritize and what they're skeptical of

    Mark one as recommended with rationale. The personas should be meaningfully different, not variations on the same viewpoint.

  3. Assess complexity - Is this a full-framework situation (complex, high-stakes, ambiguous, multi-domain) or would analysis:quick suffice?

    Use quick mode when:

    • Decision is easily reversible
    • Stakes are contained to one project or area
    • User has direct experience with this type of decision
    • Time-constrained

    Use full framework when:

    • Decision is hard to reverse
    • Multiple stakeholders or domains affected
    • Operating outside familiar territory
    • The "obvious" answer feels too easy
  4. Recommend the sequence - Which commands, in what order?

Output

First, present the personas in a clear format:

code
## Suggested Personas

### 1. [Role Title] (Recommended)
[Full persona description: 3-4 sentences covering experience, context,
formative experiences, and what shapes their judgment]

**Why this lens:** [1 sentence on why this perspective fits the subject]

### 2. [Role Title]
[Full persona description]

**Why this lens:** [1 sentence]

### 3. [Role Title] (if warranted)
[Full persona description]

**Why this lens:** [1 sentence]

Then output the exact commands with the full recommended persona embedded.

If adjacent domains were identified, include a cross-domain flag on the critique command:

code
/analysis:research [topic]. Approach as [full persona description].
/analysis:evaluate [topic]. Approach as [full persona description].
/analysis:critique [topic]. Approach as [full persona description]. Also flag concerns from a [adjacent domain expert].
/analysis:synthesize

If no significant adjacent domains, use standard critique:

code
/analysis:research [topic]. Approach as [full persona description].
/analysis:evaluate [topic]. Approach as [full persona description].
/analysis:critique [topic]. Approach as [full persona description].
/analysis:synthesize

If using quick mode:

code
/analysis:quick [topic]. Approach as [full persona description].

The user can copy these directly or substitute an alternative persona from the list.

Don't run anything yet—just give me the plan and the commands.

Available Commands

  • analysis:research – Gather external context, industry trends, evidence base
  • analysis:evaluate – Five perspectives: Informed, Steelman, Skeptic, Practical, Contrarian
  • analysis:critique – Adversarial: find fatal flaws, assumptions, failure modes. Supports optional cross-domain flag: "Also flag concerns from a [domain expert]"
  • analysis:synthesize – Structured output: summary, recommendation, tradeoffs, open questions
  • analysis:decide – Force commitment: state the decision, confidence, reversal triggers, and next action
  • analysis:quick – Lightweight mode: compressed analysis in a single pass

Subject: $ARGUMENTS