v1.3.03 loop iterations

Ceo Mode

Founders think in spirals, not straight lines — and by the time they've talked through a decision, they've moved on without documenting the conclusion. CEO Mode is a chief-of-staff-in-a-box that captures the decision, the conviction level, and the bits that are still open.

What this skill does

Founder brain dumps have a specific shape that breaks normal summarisation. The same decision gets revisited three times in one session, sometimes reaching different conclusions. Tactical to-dos ("ship by Friday") and existential strategy ("should we enter enterprise next year?") show up in the same sentence. "We should probably do X" is a preference, not a decision — but most summarisation tools promote it to a decision anyway. The result: founders end up acting on things they didn't agree to and forgetting things they did.

The skill separates four things every time: decisions actually made (HIGH conviction, with reasoning), directions leaning (MEDIUM conviction, with the evidence that would confirm or reverse), open questions (the things you raised without answering), and ideas worth exploring (LOW conviction, captured so they don't evaporate). Reversed decisions get the latest position with a note on the reversal — never silently overwritten. Implicit priorities get surfaced from how much time and energy you spent on each topic, because the words count and the emotional weight count.

The chief-of-staff lens means analysis, not just organisation. A secretary takes notes; a chief of staff names the tensions you're holding without seeing. The skill flags contradictions explicitly — "you said stay bootstrapped but also described a feature set that needs a team of 15" — because the trade-offs that define strategy are the ones founders most often refuse to look at. It won't add unsolicited strategic advice; if you ask for it, the advice goes in a clearly labelled "Advisory Notes" section so your thinking and the commentary never get blurred.

It scales to input. A three-paragraph dump gets a three-row table, not a fifteen-section brief. A 2,000-word session gets the full format. And if the dump is mostly emotional processing — repeated "I don't know", overwhelm, fatigue language — the skill says so in the processing notes and refuses to promote a fatigue-induced "maybe we should pivot" into a strategic direction. The output ends with a quick-review prompt for the founder: scan the Decisions table, anything there that's not actually decided?

When this triggers

·You just talked through your business for 20 minutes and need it sorted into decisions vs musings
·You keep re-litigating the same strategic question because no one wrote down the last answer
·You want a brief that separates decided from leaning from open
·You're mixing tactical to-dos and existential strategy in the same brain dump and need them split
·You want the tensions in your thinking surfaced, not smoothed over

Example

Trigger

User pastes a 1,400-word voice memo transcript ranging over hiring, pricing, and whether to raise.

Output

# CEO Mode Brief Topics: hiring (snr eng), pricing v2, fundraise decision. Decisions (HIGH conviction) 1. Hire snr eng before any new feature work. Reasoning: capacity is the bottleneck on every other plan. Leaning (MEDIUM — needs validation) 1. Stay bootstrapped through end of year. Would confirm: Q3 revenue clears £40k MRR. Would reverse: enterprise pilot signs and demands SLA. Open questions 1. What's the actual ceiling on self-serve pricing? Why it matters: blocks the hire ROI case. Next step: pull current cohort retention by tier. Strategic priorities (from emphasis): 1) capacity 2) pricing 3) fundraise (talked about least, may indicate not actually primary). Tensions identified: · Bootstrap stance vs. timelines that imply a team of 8 · "Speed matters" vs. "we need polish before launch"

Get this skill + 3 more

Get the full Voice-First pillar (4 skills) or the complete library.

Get the full stack — $299

What you get

188-line SKILL.md, ready to drop into ~/.claude/skills/
Tested through 3 Karpathy-loop iterations (versions v1.0.0 → v1.3.0)
Triggers automatically when relevant — no command to remember
Lifetime updates as the skill is refined further