analyzelisted
Install: claude install-skill tomcounsell/ai
# Analyze
Deliver a structured strategic assessment — convergence, tensions, overlooked factors, a direct recommendation, and one first move — on a decision, idea, or plan with meaningful stakes. Do not run it for simple factual questions, quick lookups, or casual requests without meaningful stakes.
## Framing
Compose a complete brief from what you already know: the user's raw question, relevant conversation context (constraints, goals, numbers, prior decisions), workspace files already in context that bear on the question, and what's at stake. Do not ask the user for anything.
## Run
**If a `strategic-analyst` agent type is available** (listed among this session's agent types), delegate the whole protocol to it and relay the report path it returns:
```
Agent(subagent_type="strategic-analyst", prompt="[the framed question]")
```
**Otherwise, run the same protocol yourself** with general-purpose agents:
1. **Five lenses in parallel** — spawn 5 agents in a single message, each applying one lens to the framed question (150-300 words, no hedging):
- *Downside* — failure points, fatal flaws; assumes a critical weakness exists and hunts for it
- *Foundational* — challenges the question's assumptions; may conclude the wrong question is being asked
- *Upside* — overlooked opportunities and undervalued assets, unbounded by risk
- *Outside* — zero domain context; catches what experts assume is obvious
- *Execution* — "what do you do Monday morning?"; flags ideas