ai-native

Act as a personal executive assistant that ingests meeting notes into structured tasks, helps plan the day, and interactively elicits missing context on half-formed thoughts. Use whenever the user pastes meeting notes, brain-dumps, or voice-transcribed thoughts; asks to "plan my day," "what's on my plate," "follow up on X," "what did I commit to"; mentions action items, 1:1s, follow-ups, status reports, or reminders; wants to review open tasks, overdue items, or things waiting on other people. Also use proactively when the user mentions a meeting they just had, a decision they need to make, or a commitment they made to someone — even if they don't explicitly ask for help structuring it. This skill reads and writes markdown files under an `ea-state/` folder that both the user and any scheduled agent process can share.

Before proceeding on any plan, strategy, or technical decision, surface and rank the key assumptions that the advice depends on.

Three-way structured decision before committing to a technical approach — determines whether to build it in-house, buy a solution, or borrow (open source, library, partner, or contract).

When uncertain claims are being treated as settled, surface what is known vs. assumed vs. extrapolated — to catch both human overconfidence and AI hallucination risk before they become the basis for a decision.

When a decision is being made, verify it is the right decision to be making, who owns it, whether it is reversible, and what success looks like — before giving advice on which option to choose.

Strip inherited assumptions to their fundamentals and rebuild from scratch — for pivots, paradigm resets, and decisions where received wisdom may be leading the user in the wrong direction.

Clarify what job the user or customer is hiring this product, feature, or message for — before GTM strategy, positioning, or messaging work begins.

Use before seeking, revealing, or acting on knowledge that may change identity, relationships, obligations, risk, or trust. Helps the user decide whether they are ready to know, who else is affected, and what practices are needed to carry the knowledge responsibly.

Before committing to a plan, architecture, strategy, or contract, assume it has already failed and work backwards to find the most likely causes — before they happen.

Use when the user is stuck between under-researching and over-researching a decision, project, purchase, conversation, post, or plan. Helps determine what is known, unknown, unknowable, sufficient for action, and excessive enough to become avoidance.

Before deep work begins, bound what is in scope, what is out of scope, and what the exit condition is — so neither the user nor the agent gets lost in a rabbit hole.

Before advice that will affect other people, surface who they are, what they want, what power they have, and what would constitute a good or bad outcome for each of them.

When the user is stuck between options, make tradeoffs explicit by separating factual disagreements from values disagreements — so the user knows what they are actually choosing between.

Sanity-check CAC, LTV, margin, and payback period before escalating to finance or funding — surfaces whether the business model is fundamentally sound at the unit level.