# Claude Opus 5.5 for Product Managers: 39 Claude Skills to Run the Whole Product Cycle From the project setup to Friday's wrap — 39 skills that run the product cycle in seven phases, each handing a named artifact to the next, with every figure sourced and every call left to a person. - Page: https://lantern.meet-polar-bear.com/skills/opus-for-product-managers-pack - Download: https://lantern.meet-polar-bear.com/skills/opus-for-product-managers-pack.zip - Source: https://github.com/polar-bear-org/claude-skills/tree/main/packs/opus-for-product-managers-pack - Price: free ## Install in Claude Code ``` /plugin marketplace add polar-bear-org/claude-skills /plugin install opus-for-product-managers-pack@polar-bear-skills ``` ## The 39 skills you get The set follows the product cycle, in seven phases: setting Claude up and connecting it, framing the question, discovering the evidence, deciding, building and shipping, communicating the call, and running the week. ### 1 · Set up and connect - **Set Up the Product Project** — use when: Every chat starts with “so, our product is...” and answers stay generic. Output: Project instructions, knowledge list (load, link, keep out), five test prompts. - **Write the Product Context File** — use when: Claude writes for users you do not have, or remembers last quarter's priorities. Output: Dated context file with owners, memory rules, Topics audit. - **Turn a Repeat Task Into a Skill** — use when: You paste the same long prompt every Friday and it still drifts. Output: Skill draft, three test inputs from past work, before and after check. - **Plan Your Connectors** — use when: You want Claude to read your tools but not change them by surprise. Output: Stack map, read and write actions, writes kept off, setup order, first-run tests. - **Schedule a Routine Safely** — use when: You want the Friday work to run without you, and never post as you. Output: Scheduled task spec, connectors kept and removed, draft-only rule, first-run review. ### 2 · Frame - **Frame the Problem** — use when: A leader asks to “fix retention” and nobody knows what decision is due. Output: One testable question, the decision it feeds, owner, scope, criteria, deadline. - **Read the Product's Health** — use when: You are new to the product or about to plan. Output: Sourced, dated read of usage, quality, revenue signals and feedback, unknowns. - **Map the Stakeholders** — use when: Last week's sign-off was reopened by someone you never mapped. Output: Who decides, influences and can block, by role, and how to involve each. - **Build the Issue Tree** — use when: The question is big and the team argues about where to start. Output: Sub-questions that do not overlap and cover the question, data per branch. - **Draft the Hypotheses** — use when: Discovery has no shape and data piles up without a point. Output: Answer-first guess per branch, what would kill it, test, ranked list. - **Plan the Discovery** — use when: You have two weeks between sprints to learn enough to decide. Output: Time-boxed workplan, method and owner per hypothesis, decision checkpoint. ### 3 · Discover - **Synthesise Customer Calls** — use when: Twelve recorded calls sit in the notes tool and the spec is due. Output: Themes with linked quotes, contradictions, counts by segment, next questions. - **Map the Competitors** — use when: You hear about a competitor launch from sales, a week late. Output: Sourced, dated landscape table, where you differ, optional weekly watch. - **Size the Market** — use when: Leadership asks “how big is this?” and any number becomes the plan. Output: Top-down and bottom-up ranges, sourced assumptions, the gap explained. - **Segment Customers by Needs** — use when: “Our users” means four groups and the roadmap serves none well. Output: Segments by job and circumstance, size signal, current alternative, needs. - **Analyse the Usage Data** — use when: You wait three days to learn whether a feature is used. Output: Answer with query, definitions, cleaning log, charts, caveats. - **Run a Deep Research Brief** — use when: You need the market read by Thursday, with claims you can trace. Output: Research brief, cited report, confidence per finding, still-unknown list. ### 4 · Decide - **Generate Distinct Options** — use when: The only options are the CEO's idea and a smaller version of it. Output: Three to five options that differ in kind, do nothing included, bets and costs. - **Model the Scenarios** — use when: One forecast line is in the plan and everyone knows it is wrong. Output: Base, upside and downside per option, drivers, early signals. - **Score the Options** — use when: The decision meeting turns into who argues loudest. Output: Weighted criteria agreed first, evidence per score, weight sensitivity. - **Build the Business Case** — use when: Finance asks for the return and the numbers are visible guesses. Output: One-page case, confirmed figures apart from assumptions, ranges. - **Red-Team the Plan** — use when: The plan looks finished and nobody has argued with it yet. Output: Case against, pre-mortem story, counter-evidence, what to change now. - **Write the Recommendation** — use when: You have the analysis and need a call, not another discussion. Output: Recommendation first, reasons, rejected options, risks, the ask, owner. ### 5 · Build and ship - **Sequence the Roadmap** — use when: The recommendation is approved and everything wants to be “now”. Output: Now, next, later board, cost of delay order, not-on-the-roadmap list, exec cut. - **Write the Spec in Claude Docs** — use when: The spec lives in five chats and engineers read none of them. Output: Answer-first spec, testable requirements, open questions with owners. - **Prototype in Claude Design** — use when: Leadership saw a flashy demo and expects yours by Friday. Output: Clickable prototype, layout options, edge states, test brief. - **Set Up Claude Code as a PM** — use when: You need to know how a feature really behaves before you promise. Output: Product folder, read-only setup, feature explainer with file references. - **Define the KPIs** — use when: The feature ships next month and nobody has defined success. Output: Outcome metric, leading and lagging indicators, baseline, target, guardrails. - **Read the Experiment Result** — use when: The test ended and everyone reads the dashboard differently. Output: Readout with interval, guardrails, sample ratio check, ship, iterate or stop. ### 6 · Communicate - **Write the Executive Summary** — use when: Your work is strong and invisible past the first screen. Output: One page, point first, three supported points, the ask, the risk. - **Write the Leadership Memo** — use when: A big call needs more than bullets and basics keep being relitigated. Output: Narrative pre-read for one decision meeting, read in silence first. - **Build the Deck in Claude Slides** — use when: The review is tomorrow and the doc is already done. Output: Action-title storyline, deck, speaker notes, numbers matched to the doc. - **Prepare the Hard Questions** — use when: The review is in two days and the pet feature will come up. Output: Questions by role, short answers with evidence, numbers to have ready. - **Run the Decision Meeting** — use when: Meetings end with “let's take this offline” and nothing is decided. Output: One-decision agenda, DACI roles, pre-read, decision record after. ### 7 · Run the week - **Build the Morning Brief** — use when: The first hour goes on opening twelve tabs. Output: Daily brief of what needs you, a “since date” mode after time off. - **Triage This Week's Feedback** — use when: Feedback sits in four tools until a customer escalates. Output: Tagged feedback log, new against known themes, reply drafts. - **Prepare the Weekly Update** — use when: Two hours every Friday on an update people skim in seconds. Output: Friday draft, headline, at risk, asks, one cut per audience. - **Prep for the Meeting** — use when: Back-to-back meetings and you walk into the next one cold. Output: One-page brief, decision wanted, what each role cares about, open actions. - **Wrap the Week** — use when: The week ends and Monday starts from zero again. Output: Done against planned, decisions to log, open loops, context file edits. ## Workflows - **Set up first** (Before the first analysis, so answers stop being about a generic product.) Set Up the Product Project → Write the Product Context File → Plan Your Connectors - **Frame the question** (A problem that needs one testable question.) Frame the Problem → Build the Issue Tree → Draft the Hypotheses → Plan the Discovery - **Gather the evidence** (You need themes, competitors, size, segments and usage before the call.) Synthesise Customer Calls → Map the Competitors → Size the Market → Segment Customers by Needs → Analyse the Usage Data - **Make the call** (The evidence is in and a recommendation is due.) Generate Distinct Options → Score the Options → Red-Team the Plan → Write the Recommendation - **Build it** (The call is made and the work has to be specced and shipped.) Sequence the Roadmap → Write the Spec in Claude Docs → Prototype in Claude Design → Define the KPIs - **Carry it into the room** (The decision has to be made, and defended, in a meeting.) Write the Executive Summary → Build the Deck in Claude Slides → Run the Decision Meeting - **Close the week** (Friday, and what changed has to go back into the context file.) Wrap the Week → Write the Product Context File ## The quality bar - Answer first, method named: the answer is on the first line, and every skill names the method it applies, so you can check that it was applied - No invented facts: a missing figure is marked unknown, confirmed figures are kept apart from assumptions, and every figure carries its source and date - Evidence you can open: every customer quote links to the call or ticket it came from, and each draft ends with a check — numbers traced, quotes linked, sources opened, padding cut - Work is scored, people are not: options are scored on criteria agreed before any scoring, and customer personal data never goes into memory, the context file or a deck - Read only, draft only: connectors stay read-only unless you turn one write on for one task, and scheduled routines draft and never send - A named role decides: the decision and the send stay with a named role, and legal or regulatory questions go to a qualified adviser instead of being answered - The red line: Claude reads your tools, drafts the work and runs the routines — you talk to the customers, make every call, and press send ## Questions people ask ### What is in the Claude Opus 5.5 for Product Managers pack? 39 Claude skills that follow the product cycle in seven phases: set up and connect, frame, discover, decide, build and ship, communicate, and run the week. Each skill teaches one named method, with a fixed output template and the next skill to run. ### How do I install these Claude skills? In Claude Code, two lines install all 39 at once. In Claude.ai or Claude Desktop, download the zip, open Settings, Skills, Add, and upload the skills you want. Either way it takes about five minutes. ### Do I need Claude Code to use them? No. The skills work in Claude.ai, Claude Desktop and Claude Code. In the app you upload them yourself; in Claude Code one command installs the whole pack. ### Do the skills work with ChatGPT or other assistants? They are built as Claude Skills, and installing them is a Claude feature. Each one is a plain Markdown file, so you can read it and reuse the method elsewhere, but it will not load and trigger by itself outside Claude. ### Will Claude make the product calls for me? No. Claude reads your tools, drafts the work and runs the routines. You talk to the customers, make every call, and press send. Connectors read your stack with the write actions kept off. ### Is the pack free? Yes. It is free to use inside your company and in client work, and not for resale. There is no signup: the download and the install command are on this page. ## About Written by Polar Bear (https://meet-polar-bear.com/). Published on Lantern (https://lantern.meet-polar-bear.com/), which turns your notes, documents and call transcripts into social post drafts you review and publish.