kalyvask

Drafts and critiques business pitches, investor decks, mission statements, and personal pitches using the 7-part pitch framework, S.H.I.T., and the six-word test. Use when the user is writing a VC pitch, a deck narrative, a "tell me about yourself," a LinkedIn bio, a one-pager, a founding story, or any pitch about themselves or their company. Triggers on "pitch," "deck," "investor," "mission statement," "tell me about yourself," "founding story," "LinkedIn bio," "elevator pitch."

Drafts and critiques speeches, talks, keynotes, lightning talks, demo-day pitches, all-hands updates, conference presentations, wedding toasts, and any spoken-delivery piece using Glenn Kramon's 12 wowing-the-crowd rules from Stanford GSB's Winning Writing. Use when the user is preparing a talk, presentation, keynote, panel opener, board presentation, all-hands, town hall, demo-day pitch, toast, eulogy, commencement address, or any piece that will be delivered aloud. Triggers on phrases like "I'm giving a talk," "preparing a speech," "keynote," "all-hands," "demo day pitch," "wedding toast," "panel opener," "TED-style," "deliver this aloud," "speaking at," "presentation script."

Reorganizes a memo, email, status update, or report so the bottom line is up front. Implements Kramon's BLUF (Bottom Line Up Front) rule. Use when a draft buries the lede, opens with context instead of conclusion, or builds up to the point with "organ music." Triggers on "BLUF," "lede," "buried," "memo," "status update," "exec summary," "TL;DR."

Cuts a draft to a target word count without losing substance. Two modes behind one skill. target-count hits a specific number (200 for a cold email, 500 for an op-ed, 6 for a product summary). redundancy catches the specific failure where a phrase says what the verb or context already implied — "going forward" after a future-tense verb, "as I mentioned earlier," "reduce so they are smaller." Implements Kramon's "rewrite this in less than 100 words" exercise, the six-word summary drill, and the in-class redundancy exercise. Use when the user says "shorten this," "cut this," "tighten," "make it concise," "too many words," "going forward," "say it twice," "redundant." Pass --mode target-count|redundancy|both (default both) and --target-words N for target-count.

Mode-aware writing critic. Distilled from Rachel Konrad's Stanford GSB "Winning Writing" course (spring 2026), the 10 rules of cold outreach, Heidi Roizen's mailing rules, and live class critiques. Loads a rule catalog (rules/catalog.json) and a mode definition (rules/modes.json) at invocation. Use whenever drafting or critiquing a cold email, outreach message, LinkedIn intro, exec memo, personal essay, bio, README, or LinkedIn post. Pass a mode argument to filter which rules apply.

Fixes two sentence-rhythm tells that mark prose as AI-generated. Two targets behind one skill, picked via the --target arg. fragment-chain (three or more short sentences back-to-back) and uniform-length (two or more consecutive sentences within ~2 words of each other, especially with parallel structure). Use when the user says "reads like AI," "too choppy," "same length sentences back to back," "staccato," "fragments piled up," "rhythm is off," "uniform pacing." Pass --target fragment-chain|uniform-length|all (default all).

Rewrites blunt downward or peer feedback as "what I like plus what I would like," focused on the work rather than the person, ideally phrased as a question. Use whenever the user is drafting a performance review, a Slack message to a report, a 1:1 talking point, an HR conversation script, a peer-review comment, or any line where they're tempted to criticize a colleague directly. Triggers on "give feedback to," "tell my report," "perf review line," "stop doing X," "you don't / you're not / you should stop," "smile more," "speak up more," "interrupt less," "less blunt," "burning out," "running late."

Finds the unexpected, dry, or self-deprecating angle that makes an email memorable. Turns "professional and polite" into "this person sounds like a real human." Use when a draft is technically correct but forgettable, when the user wants humor without slapstick, or when they need a subject line or sign-off with personality. Triggers on "make it funny," "more personality," "less corporate," "humor," "memorable subject line," "sign-off ideas."

Coach a founder to get their first ~10 paying customers — the tactics of early founder-led sales, not research interviews. Starts from where the buyer actually spends their day, works the warm-network → unscalable → tools ladder, and writes outreach with one clear ask. Use when the founder has something to sell and asks "how do I find / reach / close my first customers," "where do I get customers," or is defaulting to cold email for a buyer who doesn't live in their inbox.

Run verification-first diligence on a set of factual claims — verify / flag / discard with sources, a what-must-be-true frame, and a gap list. Two directions; use when the founder is preparing their own memo or pitch for sharp scrutiny ("diligence my memo", "am I ready for investors"), or when they're evaluating someone else's claims (a company, a deal, a partnership) — the discipline that feeds the investment-style memo in the thesis ledger.

Synthesize the founder's whole journey into a 1-2 page PMF memo — insight, value hypothesis, desperation evidence, validation, metrics, the surprise, and an honest verdict. The capstone artifact you hand an investor, co-founder, or early hire. Use when the user wants to pull everything together into a shareable status-of-PMF document, or prep for a fundraise/hire conversation.

Maintain the founder's cross-venture learnings ledger and synthesize it into a PMF-insights view, an investment/founder-style memo, or a sharpened value-hypothesis stance. Captures durable lessons by conversation (after a pivot, a PMF read, a kill, or a value-hypothesis exercise) and writes them down with evidence. Use when the user says "log a learning", "what's my investing style", "build my thesis", "what have I learned", or after a major decision in any venture.

Coach a career decision for someone who wants to end up building or joining startups — which job to take, how to evaluate offers, whether to take a chief-of-staff role, when to leave. This is the OFF-SPINE career layer, separate from the 00→07 venture journey. Use when the question is "where should I work / which offer / should I take this role," NOT "how do I build my company." Full reference in career/career_strategy.md.

Write a winning cold email or outreach message — for RDI research, customer discovery, fundraising, hiring, or any cold outreach. Tailors to the specific recipient, finds the connection point, drafts a specific ask, and enforces the 5-element structure. Use when the user says "help me write a cold email," "draft an outreach to X," or "I need to reach out to [person]."

Design a cheap, fast concept test that proves whether a desperate customer would WANT a candidate solution — before building it. Picks the right format (landing page, video, concierge, follow-me-home, wizard-of-oz), defines the behavioral pass threshold, and keeps the user from drifting into an expensive build. Use when the user has a solution shape and needs to validate demand (Stage 03 / pre-Stage 05).

Track and steer a customer-discovery sprint across many interviews — surface recurring patterns, score desperation across the batch, and flag when the user is talking to needy vs. desperate customers or has stopped too early. Use when the user is in the middle of customer discovery (Stage 02), running multiple interviews, and wants to know what the data is telling them so far.

Walk a founder or early-stage PM through finding product-market fit on a specific bet. Use when the user is pre-PMF (no validated value hypothesis), has a candidate insight + audience + business model, and needs help deciding what to test next, how to interpret a failed experiment, and whether to pivot (change who/how) or restart (change what). Different from pm-value-hypothesis-tester which pressure-tests a single hypothesis statement statically; this skill runs the multi-step PMF process iteratively. Built on the Lean Startup synthesis (Rachleff, Ries, Blank, Moore, Christensen, Marks, Fitzpatrick, Cook).

Critically review the user-facing surface of a PRD, design spec, or feature proposal against behavioral and UX principles. Use when a PM has a solution in hand and wants the design layer pressure-tested — defaults, friction placement, choice architecture, information density, AI surface decisions, peak-and-end moments. Distinct from pm-red-team (strategy adversary) and pm-evaluator (rubric scoring); this skill stays at the design layer and asks whether the design works with human cognition or against it. Returns the three load-bearing design holes with specific re-writes.

Audit a status update, exec review, board email, all-hands talking point, or dashboard callout for credibility leaks before sending. Flags claims that overstate ("shipped" when 5 users are in beta), cherry-picked windows ("+15% w/w" off a holiday), vanity metrics used as validation, attribution claims with obvious uncontrolled counterfactuals, and "on track" forecasts without a threshold. Use when a PM is about to communicate progress upward and wants every claim pressure-tested. Built on the principle that goodwill is a finite budget — every overclaim debits the account, and once drained the real wins stop being believed.

Pick a single North Star metric for a product. Weighs candidates across behavioral (MAU, sessions), value-delivered (messages sent, orders completed), and financial (ARR, revenue), forcing the comparison most teams skip — explainability, adoption + retention coverage, lead/lag, gameability. Use when a PM is defining a North Star, debating between MAU/DAU vs. value/revenue, or revisiting an existing one. Pushes back on jumping to a sophisticated composite or to revenue too early. Different from pm-metrics-critic, which reviews whole dashboards; this skill picks the one number.

Given a feature scope description, recommend the appropriate design-process intensity. Routes large features (M/L/XL) through the full Research → PRD → Design Concept → Detailed Design → Working Code pipeline; routes small features and bug fixes (XS/S) through an Express Lane (verbal sign-off, no Figma) that skips stages without value. Catches both over-processing of small work and under-processing of large work. Use at kickoff when scoping a new feature, or when triaging a backlog where the design team is over-loaded and you need to decide where to invest the process.

Stress-test a design, flow, or PRD section by simulating how a specific persona would walk through it step by step. Takes the design plus a target persona (defined inline or pulled from pm-state stakeholders) and outputs a walkthrough from the persona's point of view — what they see, what they think, where they get stuck, the first failure mode that would make them bounce. Designed to surface obvious failures in 10 minutes before committing engineering time or scheduling real user research. Distinct from pm-design-critic (which applies behavioral principles to the artifact) and pm-customer-interview-coach (which helps run interviews with real users); this skill simulates the user.

Load the PM agent's persistent context — personal style and preferences from `you.md`, the active project list, cross-project stakeholders, and the relevant project's state. Use as the first step in any PM workflow when you need the agent to understand the user's current work before doing anything else. The other operate-stage skills (pm-morning-brief, pm-meeting-prep, pm-meeting-debrief, pm-weekly-review, pm-stakeholder-tracker) chain to this skill first.

Pull recent Gmail threads, classify them into action categories (substantive response needed / quick reply / FYI / can ignore), and draft replies for the substantive ones. Surfaces stale threads where the user is awaiting a response and stale threads where someone is awaiting them. Used as part of pm-morning-brief or invoked directly when the inbox feels out of control. Drafts only — never auto-sends.