[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M78BWWNM9RRK8BKHTX3JFTFH":3,"skill-same:01M78BWWNM9RRK8BKHTX3JFTFH":48},{"skillNo":4,"name":5,"nameZh":6,"description":7,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":16,"categories":17,"keywords":18,"publishedAt":19,"linkOnly":20,"markdownContent":21,"markdownBody":22,"frontMatter":23,"refName":28,"commitSha":29,"skillDirectory":30,"forksCount":31,"repositoryHtmlUrl":32,"repositoryHomepageUrl":6,"ownerHtmlUrl":33,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":34,"install":35,"resources":47},"01M78BWWNM9RRK8BKHTX3JFTFH","explore-options",null,"Fan out isolated parallel perspectives on an open-ended decision, then converge on a fit-checked, skimmable shortlist. User-invoked.","adriancodes\u002Fskills","adriancodes","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F940414?v=4","01M6NJ91CRRXPC6RRKWM4CVEGJ","Adrian Martin","01M78BV1DGXZ4VG3VVC30VH0MC",1,"MIT","https:\u002F\u002Fgithub.com\u002Fadriancodes\u002Fskills\u002Fblob\u002Fmain\u002Fevals\u002Fremaining-skills-2026-10-04\u002Fsubjects\u002Foriginal\u002Fexplore-options\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: explore-options\ndescription: Fan out isolated parallel perspectives on an open-ended decision, then converge on a fit-checked, skimmable shortlist. User-invoked.\ndisable-model-invocation: true\nlicense: MIT\nmetadata:\n  category: Design\n  summary: Fans out isolated parallel frames, converges fit-first with a constraint-checked winner, and renders a 15-line shortlist that expands on demand.\n---\n\n# Explore Options\n\n## Overview\n\nObvious answers arrive first. Escape them with isolated parallel branches under distinct frames. Then converge by fit, never cleverness. Branches that see each other anchor, so keep them isolated.\n\nRender the result as a shortlist a skimmer absorbs in one glance. Hold everything else back until asked.\n\n## When to Use\n\n- Explicit invocation only: `\u002Fexplore-options \u003Cproblem>`. Never self-invoke.\n- Fits architecture, API, naming, and product decisions where an obvious answer may be costly.\n\n## Do Not Use When\n\n- One canonical answer exists: give it directly; a fan-out decorates the inevitable.\n- A bug needs a root cause: `diagnose` (if installed) owns hypothesis generation with a reproduction loop.\n- The deliverable is a module interface: `design-an-interface` (if installed) produces full competing designs; this skill produces idea-level options.\n- Requirements are unpinned: use `create-spec` when installed. Otherwise ask one direct question first.\n\n## Required Context\n\n- The problem statement and its stated constraints: scale, latency, budget, compatibility — whatever bounds a viable answer.\n- When constraints are absent, ask exactly one question for constraints and success. Do not fan out first.\n- A harness with parallel subagent dispatch (see Failure Modes when absent).\n\n## Workflow\n\n1. **Pick 5 frames.** Choose 4 matching tags and exactly 1 `wild` tag. Vary repeat runs.\n\n   Done when 5 frames are chosen, including one wild frame.\n\n2. **Diverge in parallel.** Spawn 5 isolated subagents in one message. Give each only the problem, constraints, frame prompt, and generator instruction. Never serialize them. Never share branch output. Never simulate branches in the orchestrator.\n\n   Done when all 5 branches return.\n\n   Generator instruction, verbatim per branch:\n\n   > DIVERGENT mode: generator, not critic. Produce 6 distinct ideas under this frame, one sentence each plus a one-clause rationale. The first three answers any senior engineer would give are banned — push past them. No evaluation, no ranking, no hedging. Return a JSON array only: `[{\"idea\": \"...\", \"why\": \"...\"}]`\n\n3. **Converge: fit first.** In the orchestrator, no extra agent call:\n   - Cluster the pool by underlying angle, never by surface keyword.\n   - Rank by fit to the stated problem and viability; novelty breaks ties only.\n   - Flag every attractive-but-broken idea as a trap with a one-line reason.\n   - Run the constraint check: restate the stated constraints and verify the top pick violates none. A pick that fails is demoted — it never ships as the winner, however clever.\n   - Fill the ★ wild slot with the strongest novel-but-viable idea. The wild slot never outranks a constraint-passing pick.\n\n   Done when 3 ranked picks pass and the wild slot is filled or declared empty.\n\n4. **Render the shortlist: 15 lines or fewer.** With no preamble or narration, emit exactly, in order:\n   - 1 line: the problem and its binding constraint.\n   - 3 lines: ranked picks — name, mechanism, why it fits, one line each.\n   - 1 line: ★ wild pick, marked as such.\n   - Up to 4 lines: traps, one line each with the reason.\n   - 1 line: `More: \"map\" for all clusters · \"deepen \u003Cpick>\" for a build sketch.`\n\n   No headers, no score chips, no prose between entries. Done when the reply is 15 lines or fewer and every line carries a decision-relevant fact.\n\n5. **Expand on demand only.**\n   - `map`: render every cluster — one header line per cluster angle, one line per idea. No commentary.\n   - `deepen \u003Cpick>`: call one subagent. Return a 4–8 sentence sketch, key risk, first step, and 3 child ideas.\n\n   Done when the requested layer is rendered and nothing beyond it.\n\n## Frames\n\nPick 5 per run: 4 matching the problem's tags + 1 wild.\n\n| Frame | Vantage prompt | Tags |\n|---|---|---|\n| **inversion** | Ask the opposite question: if the goal is X, how would X be guaranteed to fail? Negate each answer back into an idea. | code, design |\n| **adversary** | Attack the obvious solution — exploit it, break it, abuse the legal-but-unintended path. Invert each attack into a design idea. | code, design |\n| **transplant** | Pick one distant domain — immune systems, markets, logistics, game design — and force-fit its core mechanism onto this problem literally. | wild |\n| **extremes** | Design both ends: the $0, one-hour crudest thing that still does the load-bearing job, and the infinite-budget decade version. The range exposes the middle. | code, general |\n| **remove the assumption** | Name the one thing every answer treats as fixed — the framework, the database, request-response itself. Delete it. What becomes possible? | code, design, wild |\n| **operator at 3am** | Design so the on-call engineer never gets paged and the auditor can prove what happened. Observability, reversibility, provability first. | code, design |\n| **beginner** | A curious 10-year-old who has never seen software: naive, convention-free approaches. Ignore how it is usually done. | general, wild |\n| **hardware** | Think in latency budgets, memory layout, physical constraints. Re-ask this as a hardware or systems problem: what do the bus, cache, and clock say? | code |\n\n## Success Criteria\n\n- The shortlist is 15 lines or fewer and every line carries a decision-relevant fact\n- All 3 ranked picks pass the constraint check against the stated constraints\n- The ★ wild slot holds a viable out-of-playbook idea or is declared empty — it never outranks a constraint-passing pick\n- Branches ran as parallel, isolated subagent calls; nothing was simulated in the orchestrator's context\n- `map` and `deepen` layers were rendered only when requested\n\n## Common Mistakes\n\n| Mistake | Fix |\n|---------|-----|\n| Simulating branches sequentially in one context | Spawn real parallel subagents; isolation is the mechanism |\n| Clever pick wins over fitting pick | Fit and viability rank; novelty breaks ties; the constraint check gates the winner |\n| Rendering the full pool unprompted | 15-line shortlist; `map` and `deepen` are opt-in layers |\n| Ten variations of one idea presented as breadth | Cluster by underlying angle; one cluster means the divergence failed — respawn the weakest 2 frames |\n| \"Here are 30 ideas, you decide\" | Converge with a real ranking; refusing to commit is a cop-out |\n\n## Failure Modes\n\n- **No parallel subagents:** say so. Offer one strong prompt for 10 clustered approaches with traps. Label it anchored, not isolated.\n- **All branches converge:** report that the problem is narrower than expected. Give the direct answer without padding.\n- **No constraints obtainable:** stop after one question; rank by viability alone and say the constraint check was skipped.\n\n## Provenance\n\nAdapted from ADHD (github.com\u002FUditAkhourii\u002Fadhd, MIT). It retains isolation and the generator\u002Fcritic split. This version adds fit-first ranking, a gated wild slot, and layered output. Cost: about 6 calls, plus 1 per `deepen`.\n\n## Summary\n\nFan out 5 isolated frames. Converge by fit. Render at most 15 lines. Hold deeper material until requested.\n","# Explore Options\n\n## Overview\n\nObvious answers arrive first. Escape them with isolated parallel branches under distinct frames. Then converge by fit, never cleverness. Branches that see each other anchor, so keep them isolated.\n\nRender the result as a shortlist a skimmer absorbs in one glance. Hold everything else back until asked.\n\n## When to Use\n\n- Explicit invocation only: `\u002Fexplore-options \u003Cproblem>`. Never self-invoke.\n- Fits architecture, API, naming, and product decisions where an obvious answer may be costly.\n\n## Do Not Use When\n\n- One canonical answer exists: give it directly; a fan-out decorates the inevitable.\n- A bug needs a root cause: `diagnose` (if installed) owns hypothesis generation with a reproduction loop.\n- The deliverable is a module interface: `design-an-interface` (if installed) produces full competing designs; this skill produces idea-level options.\n- Requirements are unpinned: use `create-spec` when installed. Otherwise ask one direct question first.\n\n## Required Context\n\n- The problem statement and its stated constraints: scale, latency, budget, compatibility — whatever bounds a viable answer.\n- When constraints are absent, ask exactly one question for constraints and success. Do not fan out first.\n- A harness with parallel subagent dispatch (see Failure Modes when absent).\n\n## Workflow\n\n1. **Pick 5 frames.** Choose 4 matching tags and exactly 1 `wild` tag. Vary repeat runs.\n\n   Done when 5 frames are chosen, including one wild frame.\n\n2. **Diverge in parallel.** Spawn 5 isolated subagents in one message. Give each only the problem, constraints, frame prompt, and generator instruction. Never serialize them. Never share branch output. Never simulate branches in the orchestrator.\n\n   Done when all 5 branches return.\n\n   Generator instruction, verbatim per branch:\n\n   > DIVERGENT mode: generator, not critic. Produce 6 distinct ideas under this frame, one sentence each plus a one-clause rationale. The first three answers any senior engineer would give are banned — push past them. No evaluation, no ranking, no hedging. Return a JSON array only: `[{\"idea\": \"...\", \"why\": \"...\"}]`\n\n3. **Converge: fit first.** In the orchestrator, no extra agent call:\n   - Cluster the pool by underlying angle, never by surface keyword.\n   - Rank by fit to the stated problem and viability; novelty breaks ties only.\n   - Flag every attractive-but-broken idea as a trap with a one-line reason.\n   - Run the constraint check: restate the stated constraints and verify the top pick violates none. A pick that fails is demoted — it never ships as the winner, however clever.\n   - Fill the ★ wild slot with the strongest novel-but-viable idea. The wild slot never outranks a constraint-passing pick.\n\n   Done when 3 ranked picks pass and the wild slot is filled or declared empty.\n\n4. **Render the shortlist: 15 lines or fewer.** With no preamble or narration, emit exactly, in order:\n   - 1 line: the problem and its binding constraint.\n   - 3 lines: ranked picks — name, mechanism, why it fits, one line each.\n   - 1 line: ★ wild pick, marked as such.\n   - Up to 4 lines: traps, one line each with the reason.\n   - 1 line: `More: \"map\" for all clusters · \"deepen \u003Cpick>\" for a build sketch.`\n\n   No headers, no score chips, no prose between entries. Done when the reply is 15 lines or fewer and every line carries a decision-relevant fact.\n\n5. **Expand on demand only.**\n   - `map`: render every cluster — one header line per cluster angle, one line per idea. No commentary.\n   - `deepen \u003Cpick>`: call one subagent. Return a 4–8 sentence sketch, key risk, first step, and 3 child ideas.\n\n   Done when the requested layer is rendered and nothing beyond it.\n\n## Frames\n\nPick 5 per run: 4 matching the problem's tags + 1 wild.\n\n| Frame | Vantage prompt | Tags |\n|---|---|---|\n| **inversion** | Ask the opposite question: if the goal is X, how would X be guaranteed to fail? Negate each answer back into an idea. | code, design |\n| **adversary** | Attack the obvious solution — exploit it, break it, abuse the legal-but-unintended path. Invert each attack into a design idea. | code, design |\n| **transplant** | Pick one distant domain — immune systems, markets, logistics, game design — and force-fit its core mechanism onto this problem literally. | wild |\n| **extremes** | Design both ends: the $0, one-hour crudest thing that still does the load-bearing job, and the infinite-budget decade version. The range exposes the middle. | code, general |\n| **remove the assumption** | Name the one thing every answer treats as fixed — the framework, the database, request-response itself. Delete it. What becomes possible? | code, design, wild |\n| **operator at 3am** | Design so the on-call engineer never gets paged and the auditor can prove what happened. Observability, reversibility, provability first. | code, design |\n| **beginner** | A curious 10-year-old who has never seen software: naive, convention-free approaches. Ignore how it is usually done. | general, wild |\n| **hardware** | Think in latency budgets, memory layout, physical constraints. Re-ask this as a hardware or systems problem: what do the bus, cache, and clock say? | code |\n\n## Success Criteria\n\n- The shortlist is 15 lines or fewer and every line carries a decision-relevant fact\n- All 3 ranked picks pass the constraint check against the stated constraints\n- The ★ wild slot holds a viable out-of-playbook idea or is declared empty — it never outranks a constraint-passing pick\n- Branches ran as parallel, isolated subagent calls; nothing was simulated in the orchestrator's context\n- `map` and `deepen` layers were rendered only when requested\n\n## Common Mistakes\n\n| Mistake | Fix |\n|---------|-----|\n| Simulating branches sequentially in one context | Spawn real parallel subagents; isolation is the mechanism |\n| Clever pick wins over fitting pick | Fit and viability rank; novelty breaks ties; the constraint check gates the winner |\n| Rendering the full pool unprompted | 15-line shortlist; `map` and `deepen` are opt-in layers |\n| Ten variations of one idea presented as breadth | Cluster by underlying angle; one cluster means the divergence failed — respawn the weakest 2 frames |\n| \"Here are 30 ideas, you decide\" | Converge with a real ranking; refusing to commit is a cop-out |\n\n## Failure Modes\n\n- **No parallel subagents:** say so. Offer one strong prompt for 10 clustered approaches with traps. Label it anchored, not isolated.\n- **All branches converge:** report that the problem is narrower than expected. Give the direct answer without padding.\n- **No constraints obtainable:** stop after one question; rank by viability alone and say the constraint check was skipped.\n\n## Provenance\n\nAdapted from ADHD (github.com\u002FUditAkhourii\u002Fadhd, MIT). It retains isolation and the generator\u002Fcritic split. This version adds fit-first ranking, a gated wild slot, and layered output. Cost: about 6 calls, plus 1 per `deepen`.\n\n## Summary\n\nFan out 5 isolated frames. Converge by fit. Render at most 15 lines. Hold deeper material until requested.\n",{"name":5,"license":15,"metadata":24,"description":7,"disable-model-invocation":27},{"summary":25,"category":26},"Fans out isolated parallel frames, converges fit-first with a constraint-checked winner, and renders a 15-line shortlist that expands on demand.","Design",true,"main","9a3fe345285b86c3071deb2ecb302dff99a855b8","evals\u002Fremaining-skills-2026-10-04\u002Fsubjects\u002Foriginal\u002Fexplore-options",0,"https:\u002F\u002Fgithub.com\u002Fadriancodes\u002Fskills","https:\u002F\u002Fgithub.com\u002Fadriancodes","2026-11-12 06:06:49",{"source":8,"skillName":5,"commands":36,"prompt":46,"zipAvailable":27},[37,40,43],{"tool":38,"command":39},"npx","npx skills add adriancodes\u002Fskills --skill explore-options",{"tool":41,"command":42},"bunx","bunx skills add adriancodes\u002Fskills --skill explore-options",{"tool":44,"command":45},"pnpm","pnpm dlx skills add adriancodes\u002Fskills --skill explore-options","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fadriancodes\u002Fskills\u002Fblob\u002Fmain\u002Fevals\u002Fremaining-skills-2026-10-04\u002Fsubjects\u002Foriginal\u002Fexplore-options\u002FSKILL.md",[],[49,56,63,70,77,84],{"skillNo":50,"name":51,"nameZh":6,"description":52,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":53,"categories":54,"keywords":55,"publishedAt":19},"01M78C1QNZAXX06QMDJ8R65TTV","write-agent-instructions","Make the smallest evidence-backed change to persistent project instructions and test downstream behavior.","https:\u002F\u002Fgithub.com\u002Fadriancodes\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fwrite-agent-instructions\u002FSKILL.md",[],[],{"skillNo":57,"name":58,"nameZh":6,"description":59,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":60,"categories":61,"keywords":62,"publishedAt":19},"01M78C1K3QVP67S0TRERZ5YNGB","verify-work","Use when a finished, runnable artifact (a script, config, parser, prompt, or CLI, not a diff under review) must survive hostile input before shipping: the user asks to \"verify this script\", \"find the loopholes\", \"test this properly against bad input\", or asks \"is this ready to ship\" about one. Also when a shipped artifact failed on an input nobody tried, or when verification so far is a happy-path smoke test.","https:\u002F\u002Fgithub.com\u002Fadriancodes\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fverify-work\u002FSKILL.md",[],[],{"skillNo":64,"name":65,"nameZh":6,"description":66,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":67,"categories":68,"keywords":69,"publishedAt":19},"01M78C1DDYDEQ7H1HNW84VW1V3","understand-codebase","Use when the user asks to \"understand this codebase\", \"how does this feature work in this repo?\", \"trace this feature through the code\", or needs onboarding, an architecture explanation, or answers about unfamiliar code. Also use for read-only code-path questions before a change.","https:\u002F\u002Fgithub.com\u002Fadriancodes\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Funderstand-codebase\u002FSKILL.md",[],[],{"skillNo":71,"name":72,"nameZh":6,"description":73,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":74,"categories":75,"keywords":76,"publishedAt":19},"01M78C18C6TG62PY1PQBGCSKXR","tdd","Use when writing or changing production behavior: implementing a feature, fixing a bug whose cause is already known, or altering stored, transformed, migrated, serialized, or externally transmitted data. Also use when a user says \"use TDD\", \"test first\", or \"red-green-refactor\", and when code is being written before a failing test. For a failure whose cause is unknown, diagnose first.","https:\u002F\u002Fgithub.com\u002Fadriancodes\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Ftdd\u002FSKILL.md",[],[],{"skillNo":78,"name":79,"nameZh":6,"description":80,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":81,"categories":82,"keywords":83,"publishedAt":19},"01M78C151714X4Z3VQW5F23FZW","simplify-code","Use when the user asks to \"simplify this code\", \"remove the fluff\", or \"clean up this over-engineered module\", or wants working code that is bloated, duplicated, over-abstracted, or dependency-heavy reduced without changing behavior.","https:\u002F\u002Fgithub.com\u002Fadriancodes\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Fsimplify-code\u002FSKILL.md",[],[],{"skillNo":85,"name":86,"nameZh":6,"description":87,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":88,"categories":89,"keywords":90,"publishedAt":19},"01M78C127TPYNJBX0B4BCF7246","review-session","Review an agent session for evidence-backed friction and the smallest measurable remedies.","https:\u002F\u002Fgithub.com\u002Fadriancodes\u002Fskills\u002Fblob\u002Fmain\u002Fskills\u002Freview-session\u002FSKILL.md",[],[]]