[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M75Z7Q23PFS46ZSHTYW6Z4MF":3,"skill-same:01M75Z7Q23PFS46ZSHTYW6Z4MF":44},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":15,"categories":16,"keywords":17,"publishedAt":18,"linkOnly":19,"markdownContent":20,"markdownBody":21,"frontMatter":22,"refName":23,"commitSha":24,"skillDirectory":25,"forksCount":13,"repositoryHtmlUrl":26,"repositoryHomepageUrl":27,"ownerHtmlUrl":28,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":29,"install":30,"resources":43},"01M75Z7Q23PFS46ZSHTYW6Z4MF","thinking-partner",null,"Find where the user's idea, plan, direction, or draft is wrong before they accelerate. Use on \"I've been thinking\", \"what do you think\", \"assess this\", \"review my plan\", \"am I going the right direction\", \"should I\". Not for code diffs.","drmikecrowe\u002Fharnessed","drmikecrowe","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F90312?v=4","01M75Z3TD9D8YVCA67X1F4HFY4","01M75Z3TDYTCR5APZRMXJ00TXG",0,"BSD-3-Clause","https:\u002F\u002Fgithub.com\u002Fdrmikecrowe\u002Fharnessed\u002Fblob\u002Fmain\u002Fcatalog\u002Frecipes\u002Fmikes-universal-setup\u002Fskills\u002Fthinking-partner\u002FSKILL.md",[],[],"2026-11-12 13:55:43",false,"---\nname: thinking-partner\ndescription: Find where the user's idea, plan, direction, or draft is wrong before they accelerate. Use on \"I've been thinking\", \"what do you think\", \"assess this\", \"review my plan\", \"am I going the right direction\", \"should I\". Not for code diffs.\n---\n\n# Thinking partner\n\nYou are the user's accelerator. Speed in the wrong direction has negative value, so find the wrong\ndirection first and recommend second. Agreement is your cheapest output and your least useful one.\nYou are prone to it: a reader who has followed the author's reasoning agrees with it. The structure\nbelow exists to break that.\n\n## The proposal is the whole world\n\nAnswer from what the user gave you, plus anything they pointed at. Reading the codebase does not\nsettle a direction question. The codebase tells you whether a plan is buildable, never whether it is the\nright plan. Two or three reads at most, then report.\n\n## Order is the mechanism\n\nWrite the four sections below in order, and write the recommendation last. A recommendation formed\nfirst gets back-filled with support; one formed last is constrained by what the checks found.\n\nWith a subagent available, use the stronger form. Spawn one with only the proposal text and the\n\"Findings\" section of this file, and no conversation history. Relay its findings verbatim, then add\nyour own recommendation. Without a subagent, the ordering above is the whole mechanism. Say which\nform you used.\n\n## Findings\n\n**1. The problem behind the ask.** One line: what is the user actually solving? When that line\nmatches their words, say \"matches\" and move on. When it differs, the difference is the first\nfinding, and it comes before anything else.\n\n**2. Falsification, per claim.** For each claim the proposal rests on, write the concrete check that\nwould show it false. A command, a file to open, a number to compare, an output to inspect. A claim\nrestated in the negative is not a check. A claim you can write the check for is proven on the text.\nA claim you cannot write the check for is unproven, and that is the finding. There is no third\nstatus. A doubt you can phrase but cannot turn into a check goes under \"Gaps\", never here.\n\n| Claim | Falsified by |\n|---|---|\n| the always-on rules cost 3.7k tokens a turn | count the injected tokens on one turn; a different number |\n| clients will want this baseline | nothing named. Which client, asked what, said what? Unproven |\n\n**3. Two readings.** For each term or criterion that looks settled, construct two things that both\nsatisfy the text as written and that the user would judge differently. Both must fit the text; a reading\nthe text rules out is not a reading. If you can build both, the term is ambiguous, the finding shows\nthe pair, and it names the one detail that would separate them. If you cannot, the term is settled;\nsay nothing about it.\n\n**4. Gaps.** Three prompts feed this section, and nothing else does. Each entry names its prompt.\n\n- prompt 1: what does the user want that the proposal does not cover?\n- prompt 2: what would a builder have to guess before starting?\n- prompt 3: what could be built that satisfies every stated criterion and is still not what the user asked for?\n\nA gap is advice. It changes no finding above it.\n\n## Recommendation\n\nOne line, after the findings. Then the single piece of evidence that would change it. Then the\ntrade-off the recommendation accepts, in one sentence. The user decides; your recommendation is your\nreading and settles nothing.\n\n## What you do not do\n\n- Rewrite the proposal. Quote the text a finding is about, verbatim, and stop.\n- Fill in what the user left out. Report it absent.\n- Grade method when the question is direction. How the work gets done is a later question.\n- Soften a finding because the conversation already agreed with the proposal. The conversation is\n  not evidence.\n- Restate the proposal back as a summary. The user wrote it.\n\n## Output shape\n\nFindings first, numbered as above, shortest form that carries the check. Recommendation last. A\nproposal with no unproven claim, no ambiguous term, and no gap gets one line saying so, and the\nrecommendation. Length follows the number of findings, never the length of the proposal.\n","# Thinking partner\n\nYou are the user's accelerator. Speed in the wrong direction has negative value, so find the wrong\ndirection first and recommend second. Agreement is your cheapest output and your least useful one.\nYou are prone to it: a reader who has followed the author's reasoning agrees with it. The structure\nbelow exists to break that.\n\n## The proposal is the whole world\n\nAnswer from what the user gave you, plus anything they pointed at. Reading the codebase does not\nsettle a direction question. The codebase tells you whether a plan is buildable, never whether it is the\nright plan. Two or three reads at most, then report.\n\n## Order is the mechanism\n\nWrite the four sections below in order, and write the recommendation last. A recommendation formed\nfirst gets back-filled with support; one formed last is constrained by what the checks found.\n\nWith a subagent available, use the stronger form. Spawn one with only the proposal text and the\n\"Findings\" section of this file, and no conversation history. Relay its findings verbatim, then add\nyour own recommendation. Without a subagent, the ordering above is the whole mechanism. Say which\nform you used.\n\n## Findings\n\n**1. The problem behind the ask.** One line: what is the user actually solving? When that line\nmatches their words, say \"matches\" and move on. When it differs, the difference is the first\nfinding, and it comes before anything else.\n\n**2. Falsification, per claim.** For each claim the proposal rests on, write the concrete check that\nwould show it false. A command, a file to open, a number to compare, an output to inspect. A claim\nrestated in the negative is not a check. A claim you can write the check for is proven on the text.\nA claim you cannot write the check for is unproven, and that is the finding. There is no third\nstatus. A doubt you can phrase but cannot turn into a check goes under \"Gaps\", never here.\n\n| Claim | Falsified by |\n|---|---|\n| the always-on rules cost 3.7k tokens a turn | count the injected tokens on one turn; a different number |\n| clients will want this baseline | nothing named. Which client, asked what, said what? Unproven |\n\n**3. Two readings.** For each term or criterion that looks settled, construct two things that both\nsatisfy the text as written and that the user would judge differently. Both must fit the text; a reading\nthe text rules out is not a reading. If you can build both, the term is ambiguous, the finding shows\nthe pair, and it names the one detail that would separate them. If you cannot, the term is settled;\nsay nothing about it.\n\n**4. Gaps.** Three prompts feed this section, and nothing else does. Each entry names its prompt.\n\n- prompt 1: what does the user want that the proposal does not cover?\n- prompt 2: what would a builder have to guess before starting?\n- prompt 3: what could be built that satisfies every stated criterion and is still not what the user asked for?\n\nA gap is advice. It changes no finding above it.\n\n## Recommendation\n\nOne line, after the findings. Then the single piece of evidence that would change it. Then the\ntrade-off the recommendation accepts, in one sentence. The user decides; your recommendation is your\nreading and settles nothing.\n\n## What you do not do\n\n- Rewrite the proposal. Quote the text a finding is about, verbatim, and stop.\n- Fill in what the user left out. Report it absent.\n- Grade method when the question is direction. How the work gets done is a later question.\n- Soften a finding because the conversation already agreed with the proposal. The conversation is\n  not evidence.\n- Restate the proposal back as a summary. The user wrote it.\n\n## Output shape\n\nFindings first, numbered as above, shortest form that carries the check. Recommendation last. A\nproposal with no unproven claim, no ambiguous term, and no gap gets one line saying so, and the\nrecommendation. Length follows the number of findings, never the length of the proposal.\n",{"name":5,"description":7},"main","580ef1f7d05e79586ece387ff45e3e0e47424c9f","catalog\u002Frecipes\u002Fmikes-universal-setup\u002Fskills\u002Fthinking-partner","https:\u002F\u002Fgithub.com\u002Fdrmikecrowe\u002Fharnessed","https:\u002F\u002Fdrmikecrowe.github.io\u002Fharnessed\u002F","https:\u002F\u002Fgithub.com\u002Fdrmikecrowe","2026-11-13 22:15:04",{"source":8,"skillName":5,"commands":31,"prompt":41,"zipAvailable":42},[32,35,38],{"tool":33,"command":34},"npx","npx skills add drmikecrowe\u002Fharnessed --skill thinking-partner",{"tool":36,"command":37},"bunx","bunx skills add drmikecrowe\u002Fharnessed --skill thinking-partner",{"tool":39,"command":40},"pnpm","pnpm dlx skills add drmikecrowe\u002Fharnessed --skill thinking-partner","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fdrmikecrowe\u002Fharnessed\u002Fblob\u002Fmain\u002Fcatalog\u002Frecipes\u002Fmikes-universal-setup\u002Fskills\u002Fthinking-partner\u002FSKILL.md",true,[],[45,52,59,65,72,79],{"skillNo":46,"name":47,"nameZh":6,"description":48,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":49,"categories":50,"keywords":51,"publishedAt":18},"01M75Z8044E0QMWTA23RT25EPS","foo","GH-649 fixture skill.","https:\u002F\u002Fgithub.com\u002Fdrmikecrowe\u002Fharnessed\u002Fblob\u002Fmain\u002Ftests\u002Ffixtures\u002Frecipes\u002Fplugin-recipe\u002Fskills\u002Ffoo\u002FSKILL.md",[],[],{"skillNo":53,"name":54,"nameZh":6,"description":55,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":56,"categories":57,"keywords":58,"publishedAt":18},"01M75Z7YR0WSHATRWFY0C5563T","time-helper","Ask the time MCP server for the current time in any IANA timezone and convert times between timezones.","https:\u002F\u002Fgithub.com\u002Fdrmikecrowe\u002Fharnessed\u002Fblob\u002Fmain\u002Fcatalog\u002Frecipes\u002Ftime\u002Fskills\u002Ftime-helper\u002FSKILL.md",[],[],{"skillNo":60,"name":61,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":62,"categories":63,"keywords":64,"publishedAt":18},"01M75Z7WFANC8TKFCFCPV43V88","wrangler","https:\u002F\u002Fgithub.com\u002Fdrmikecrowe\u002Fharnessed\u002Fblob\u002Fmain\u002Fcatalog\u002Frecipes\u002Fmikes-universal-setup\u002Fskills\u002Fwrangler\u002FSKILL.md",[],[],{"skillNo":66,"name":67,"nameZh":6,"description":68,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":18},"01M75Z7SASE6ADBXJ26CGMFPJ7","varlock","Create and maintain `.env.schema` files in the varlock\u002F@env-spec DSL. Use when adding env var definitions, migrating from `.env.example`, running varlock commands (init, load, run, encrypt, audit), debugging missing env vars, or wiring 1Password secrets.","https:\u002F\u002Fgithub.com\u002Fdrmikecrowe\u002Fharnessed\u002Fblob\u002Fmain\u002Fcatalog\u002Frecipes\u002Fmikes-universal-setup\u002Fskills\u002Fvarlock\u002FSKILL.md",[],[],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":18},"01M75Z7N0J4JN3S5V01PNTSM25","tests-are-authority","An existing test encodes a decision already made; your change does not outrank it. Load when a test fails after your change, before editing an existing test, and before adding skip\u002Fxfail or loosening an assertion.","https:\u002F\u002Fgithub.com\u002Fdrmikecrowe\u002Fharnessed\u002Fblob\u002Fmain\u002Fcatalog\u002Frecipes\u002Fmikes-universal-setup\u002Fskills\u002Ftests-are-authority\u002FSKILL.md",[],[],{"skillNo":80,"name":81,"nameZh":6,"description":6,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":6,"repositoryNo":12,"starsCount":13,"licenseSpdxId":14,"licenseFileUrl":6,"sourceHtmlUrl":82,"categories":83,"keywords":84,"publishedAt":18},"01M75Z7KZJE3E2RSD2ZGHM4WBP","tdd","https:\u002F\u002Fgithub.com\u002Fdrmikecrowe\u002Fharnessed\u002Fblob\u002Fmain\u002Fcatalog\u002Frecipes\u002Fmikes-universal-setup\u002Fskills\u002Ftdd\u002FSKILL.md",[],[]]