Strategy check

Strategy check

Pressure-tests whether a proposed decision is on-thesis. This is a thinking tool for the CEO — read-only, no outbound actions.

Workflow

1. Pull the canon

• Company vision doc from the path set in role-ceo CLAUDE.md. If no vision doc exists, stop and say so — don't bluff a check against a missing standard.
• Current quarter's goals / OKRs from Notion.
• Last 2-3 board updates for the narrative arc leadership is telling externally.

2. Frame the decision

Restate $ARGUMENTS in one sentence. If the user's phrasing is fuzzy, sharpen it and confirm before proceeding. The check is only as good as the question.

3. Run the check

Evaluate the decision against five dimensions:

Dimension	Ask
Vision fit	Does this move us toward the stated vision, or sideways, or backward?
Current priority fit	Is this one of the 2-3 things we said we'd focus on this quarter, or a distraction?
Leverage	Does this compound (platform / team capacity / customer moat), or is it one-shot?
Reversibility	If we're wrong, can we unwind it? (Type 1 vs Type 2 in the two-way-door sense.)
Opportunity cost	What are we saying no to by saying yes?

For each dimension, give a one-line judgment and an honest concern if there is one.

4. Recommend

Pick one:

• Proceed — clear fit, recommended approach
• Proceed with changes — directionally right but reframe scope/sequencing/owner in specific ways
• Delay — right idea, wrong timing; specify trigger that would flip this to proceed
• Don't do it — off-thesis or bad cost/benefit; name the better alternative if there is one

5. Output

# Strategy check — <one-line summary of decision>

**Recommendation:** <Proceed | Proceed with changes | Delay | Don't do it>

**One-line rationale:** <why, in a sentence>

## The decision as stated
<restated cleanly>

## Check

**Vision fit:** <one line + concern>
**Current priority fit:** <one line + concern>
**Leverage:** <one line + concern>
**Reversibility:** <one line + concern>
**Opportunity cost:** <what we'd be saying no to>

## If you proceed
<2-3 bullets on how to do it well, or what would de-risk it>

## If you don't
<2-3 bullets on what to do instead, or trigger conditions for revisiting>

## Open questions
- <question that, if answered, would change the recommendation>

Don't

• Don't hedge the recommendation to be polite. The point of this skill is a strong call, weakly held.
• Don't run the check without actually reading the vision doc. "Proceed — looks aligned" without evidence is noise.
• Don't soften the opportunity cost. The thing we're saying no to is the whole point.