[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6AD75Q76YWWXY8ZQKY82PEE":3,"skill-same:01M6AD75Q76YWWXY8ZQKY82PEE":117},{"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":31,"publishedAt":56,"linkOnly":57,"markdownContent":58,"markdownBody":59,"frontMatter":60,"refName":65,"commitSha":66,"skillDirectory":67,"forksCount":68,"repositoryHtmlUrl":69,"repositoryHomepageUrl":70,"ownerHtmlUrl":71,"ownerBlogUrl":72,"ownerDescription":73,"updatedAt":74,"install":75,"resources":88},"01M6AD75Q76YWWXY8ZQKY82PEE","autoresearch",null,"Autonomous iterative experimentation loop for any programming task. Guides the user through defining goals, measurable metrics, and scope constraints, then runs an autonomous loop of code changes, testing, measuring, and keeping\u002Fdiscarding results. Inspired by Karpathy's autoresearch. USE FOR: autonomous improvement, iterative optimization, experiment loop, auto research, performance tuning, automated experimentation, hill climbing, try things automatically, optimize code, run experiments, autonomous coding loop. DO NOT USE FOR: one-shot tasks, simple bug fixes, code review, or tasks without a measurable metric.","github\u002Fawesome-copilot","github","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F9919?v=4","01M6AC83DS133TPXFSP1DHNCC6","GitHub","01M6AD3PBDYPK7NVFJNJ66Y8RS",41081,"MIT","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fautoresearch\u002FSKILL.md",[18,23,26],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000007","代码审查","SCENARIO","使用场景",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发",{"categoryNo":27,"categoryName":28,"taxonomyType":29,"taxonomyTypeDesc":30},"01JBSEEDCAT000000000000030","开发工程师","PROFESSION","适用职业",[5,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,9,53,54,55],"https","guides","scope","project","optimize","bug","auto","karpathy","autonomous","access","loop","repository","code review","metrics","git","code","run","tasks","experiments","research","performance","terminal","review","testing","2026-11-10 10:12:30",false,"---\nname: autoresearch\ndescription: 'Autonomous iterative experimentation loop for any programming task. Guides the user through defining goals, measurable metrics, and scope constraints, then runs an autonomous loop of code changes, testing, measuring, and keeping\u002Fdiscarding results. Inspired by Karpathy''s autoresearch. USE FOR: autonomous improvement, iterative optimization, experiment loop, auto research, performance tuning, automated experimentation, hill climbing, try things automatically, optimize code, run experiments, autonomous coding loop. DO NOT USE FOR: one-shot tasks, simple bug fixes, code review, or tasks without a measurable metric.'\nlicense: MIT\ncompatibility: Requires git. The project must be a git repository. Requires terminal access to run commands.\nmetadata:\n  author: luiscantero\n  inspired-by: https:\u002F\u002Fgithub.com\u002Fkarpathy\u002Fautoresearch\n---\n\n# Autoresearch: Autonomous Iterative Experimentation\n\nAn autonomous experimentation loop for any programming task. You define the goal and how to measure it; the agent iterates autonomously -- modifying code, running experiments, measuring results, and keeping or discarding changes -- until interrupted.\n\nThis skill is inspired by [Karpathy's autoresearch](https:\u002F\u002Fgithub.com\u002Fkarpathy\u002Fautoresearch), generalized from ML training to **any programming task with a measurable outcome**.\n\n---\n\n## Agent Behavior Rules\n\n1. **DO** guide the user through the Setup phase interactively before starting the loop.\n2. **DO** establish a baseline measurement before making any changes.\n3. **DO** commit every experiment attempt before running it (so it can be reverted cleanly).\n4. **DO** keep a results log (TSV) tracking every experiment.\n5. **DO** revert changes that do not improve the metric (git reset to last known good).\n6. **DO** run autonomously once the loop starts -- never pause to ask \"should I continue?\".\n7. **DO NOT** modify files the user marked as out-of-scope.\n8. **DO NOT** skip the measurement step -- every experiment must be measured.\n9. **DO NOT** keep changes that regress the metric unless the user explicitly allowed trade-offs.\n10. **DO NOT** install new dependencies or make environment changes unless the user approved it.\n\n---\n\n## Phase 1: Setup (Interactive)\n\nBefore any experimentation begins, work with the user to establish these parameters.\nAsk the user directly for each item. Do not assume or skip any.\n\n### 1.1 Define the Goal\n\nAsk the user:\n\n> **What are you trying to improve or optimize?**\n>\n> Examples: execution time, memory usage, binary size, test pass rate, code coverage,\n> API response latency, throughput, error rate, benchmark score, build time, bundle size,\n> lines of code, cyclomatic complexity, etc.\n\nRecord the user's answer as the **goal**.\n\n### 1.2 Define the Metric\n\nAsk the user:\n\n> **How do we measure success? What exact command produces the metric?**\n>\n> I need:\n> 1. **The command** to run (e.g., `dotnet test`, `npm run benchmark`, `time .\u002Fbuild.sh`, `pytest --tb=short`)\n> 2. **How to extract the metric** from the output (e.g., a regex pattern, a specific line, a JSON field)\n> 3. **Direction**: Is lower better or higher better?\n>\n> Example: \"Run `dotnet test --logger trx`, count passing tests. Higher is better.\"\n> Example: \"Run `hyperfine '.\u002Fmy-program'`, extract mean time. Lower is better.\"\n\nRecord:\n- `METRIC_COMMAND`: the command to run\n- `METRIC_EXTRACTION`: how to extract the numeric metric from output\n- `METRIC_DIRECTION`: `lower_is_better` or `higher_is_better`\n\n### 1.3 Define the Scope\n\nAsk the user:\n\n> **Which files or directories am I allowed to modify?**\n>\n> And which files are OFF LIMITS (read-only)?\n\nRecord:\n- `IN_SCOPE_FILES`: files\u002Fdirs the agent may edit\n- `OUT_OF_SCOPE_FILES`: files\u002Fdirs that must not be modified\n\n### 1.4 Define Constraints\n\nAsk the user:\n\n> **Are there any constraints I should respect?**\n>\n> Examples:\n> - Time budget per experiment (e.g., \"each run should take \u003C 2 minutes\")\n> - No new dependencies\n> - Must keep all existing tests passing\n> - Must not change the public API\n> - Must maintain backward compatibility\n> - VRAM\u002Fmemory limit\n> - Code complexity limits (prefer simpler solutions)\n\nRecord as `CONSTRAINTS`.\n\n### 1.5 Define the Experiment Budget (Optional)\n\nAsk the user:\n\n> **How many experiments should I run, or should I just keep going until you stop me?**\n>\n> You can say a number (e.g., \"try 20 experiments\") or \"unlimited\" (I'll run until you interrupt).\n\nRecord as `MAX_EXPERIMENTS` (number or `unlimited`).\n\n### 1.6 Simplicity Criterion\n\nInform the user of the default simplicity policy:\n\n> **Simplicity policy (default):** All else being equal, simpler is better. A small improvement\n> that adds ugly complexity is not worth it. Removing code while maintaining or improving\n> the metric is a great outcome. I'll weigh the complexity cost against the improvement\n> magnitude. Does this policy work for you, or do you want to adjust it?\n\nRecord any adjustments as `SIMPLICITY_POLICY`.\n\n### 1.7 Confirm Setup\n\nSummarize all parameters back to the user in a clear table:\n\n| Parameter          | Value                        |\n| ------------------ | ---------------------------- |\n| Goal               | ...                          |\n| Metric command     | ...                          |\n| Metric extraction  | ...                          |\n| Direction          | lower is better \u002F higher ... |\n| In-scope files     | ...                          |\n| Out-of-scope files | ...                          |\n| Constraints        | ...                          |\n| Max experiments    | ...                          |\n| Simplicity policy  | ...                          |\n\nAsk the user to confirm. Do not proceed until confirmed.\n\n---\n\n## Phase 2: Branch & Baseline\n\nOnce the user confirms:\n\n1. **Create a branch**: Propose a tag based on today's date (e.g., `autoresearch\u002Fmar17`).\n   Create the branch: `git checkout -b autoresearch\u002F\u003Ctag>`.\n\n2. **Read in-scope files**: Read all files that are in scope to build full context of the current state.\n\n3. **Initialize results.tsv**: Create `results.tsv` in the repo root with the header row:\n   ```\n   experiment\tcommit\tmetric\tstatus\tdescription\n   ```\n   Add `results.tsv` and `run.log` to `.git\u002Finfo\u002Fexclude` (append if not already present) so they stay untracked without modifying any tracked files.\n\n4. **Run the baseline**: Execute the metric command on the current unmodified code.\n   Record the result as experiment `0` with status `baseline` in `results.tsv`.\n\n5. **Report baseline** to the user:\n   > Baseline established: **[metric_name] = [value]**\n   > Starting autonomous experimentation loop.\n\n---\n\n## Phase 3: Experiment Loop\n\nRun this loop continuously. Do not stop to ask the user. Run until:\n- `MAX_EXPERIMENTS` is reached, OR\n- The user manually interrupts\n\n### For each experiment:\n\n```\nLOOP:\n  1. THINK   - Analyze previous results and the current code.\n               Generate an experiment hypothesis.\n               Consider: what worked, what didn't, what hasn't been tried.\n\n  2. EDIT    - Modify the in-scope file(s) to implement the idea.\n               Keep changes focused and minimal per experiment.\n\n  3. COMMIT  - git add + git commit with a short descriptive message.\n               Format: \"experiment: \u003Cshort description of what changed>\"\n\n  4. RUN     - Execute the metric command.\n               Redirect output to run.log so it does not flood the context window.\n               Use shell-appropriate redirection:\n               - Bash\u002FZsh: `\u003Ccommand> > run.log 2>&1`\n               - PowerShell: `\u003Ccommand> *> run.log`\n\n  5. MEASURE - Extract the metric from run.log.\n               If extraction fails (crash\u002Ferror), read the last 50 lines\n               of run.log for the error.\n\n  6. DECIDE  - Compare metric to the current best:\n               - IMPROVED: Keep the commit. Update the \"best\" baseline.\n                 Log status = \"keep\".\n               - SAME OR WORSE: Revert. `git reset --hard HEAD~1`.\n                 Log status = \"discard\".\n               - CRASH: Attempt a quick fix (typo, import, simple error).\n                 Amend the experiment commit (`git commit --amend`) with the fix\n                 and rerun. The experiment keeps its original number.\n                 If unfixable after 2 attempts, revert the entire experiment\n                 (`git reset --hard HEAD~1`) and log status = \"crash\".\n\n  7. LOG     - Append a row to results.tsv:\n               experiment_number  commit_hash  metric_value  status  description\n\n  8. CONTINUE - Go to step 1.\n```\n\n### Experiment Strategy\n\nWhen generating experiment ideas, follow this priority order:\n\n1. **Low-hanging fruit first**: Simple parameter tweaks, obvious inefficiencies.\n2. **Informed by results**: If a direction showed promise, explore further in that direction.\n3. **Diversify after plateaus**: If the last 3-5 experiments all failed, try a different approach entirely.\n4. **Combine winners**: If experiments A and B each improved independently, try combining them.\n5. **Simplification passes**: Periodically try removing code\u002Fcomplexity to see if the metric holds.\n6. **Radical changes**: After exhausting incremental ideas, try larger architectural changes.\n\n### Handling Constraints\n\n- **Time budget**: If a run exceeds 2x the expected duration, kill it and treat as a crash.\n- **Existing tests**: If constraints require tests to pass, run them before\u002Fafter and revert if they break.\n- **Memory\u002Fresources**: Monitor and revert if resource usage exceeds stated limits.\n\n---\n\n## Phase 4: Reporting\n\nWhen the loop ends (budget reached or user interrupts):\n\n1. **Print the full results.tsv** as a formatted table.\n2. **Summarize**:\n   - Total experiments run\n   - Experiments kept \u002F discarded \u002F crashed\n   - Starting metric (baseline) vs. final metric\n   - Improvement percentage\n   - Top 3 most impactful changes\n3. **Show the cumulative git log** of kept experiments:\n   `git log --oneline \u003Cstart_commit>..HEAD`\n4. **Recommend next steps**: Based on the results, suggest what a human researcher might try next (ideas that were too risky\u002Fcomplex for automated experimentation).\n\n---\n\n## Quick Reference\n\n### Results TSV Format\n\nTab-separated, 5 columns:\n\n```\nexperiment\tcommit\tmetric\tstatus\tdescription\n0\ta1b2c3d\t0.997900\tbaseline\tunmodified code\n1\tb2c3d4e\t0.993200\tkeep\tincrease learning rate to 0.04\n2\tc3d4e5f\t1.005000\tdiscard\tswitch to GeLU activation\n3\td4e5f6g\t0.000000\tcrash\tdouble model width (OOM)\n```\n\n### Git Workflow\n\n- All experiments happen on the `autoresearch\u002F\u003Ctag>` branch\n- Each experiment is committed before running\n- Failed experiments are reverted with `git reset --hard HEAD~1`\n- Successful experiments advance the branch\n- `results.tsv` and `run.log` stay untracked (added to `.git\u002Finfo\u002Fexclude`)\n\n### Key Principles\n\n1. **Measure everything**: No experiment without a measurement.\n2. **Revert failures**: The branch only advances on improvements.\n3. **Stay autonomous**: Never stop to ask. Think harder if stuck.\n4. **Keep it simple**: Complexity is a cost. Weigh it against gains.\n5. **Log everything**: The TSV is the research journal.\n","# Autoresearch: Autonomous Iterative Experimentation\n\nAn autonomous experimentation loop for any programming task. You define the goal and how to measure it; the agent iterates autonomously -- modifying code, running experiments, measuring results, and keeping or discarding changes -- until interrupted.\n\nThis skill is inspired by [Karpathy's autoresearch](https:\u002F\u002Fgithub.com\u002Fkarpathy\u002Fautoresearch), generalized from ML training to **any programming task with a measurable outcome**.\n\n---\n\n## Agent Behavior Rules\n\n1. **DO** guide the user through the Setup phase interactively before starting the loop.\n2. **DO** establish a baseline measurement before making any changes.\n3. **DO** commit every experiment attempt before running it (so it can be reverted cleanly).\n4. **DO** keep a results log (TSV) tracking every experiment.\n5. **DO** revert changes that do not improve the metric (git reset to last known good).\n6. **DO** run autonomously once the loop starts -- never pause to ask \"should I continue?\".\n7. **DO NOT** modify files the user marked as out-of-scope.\n8. **DO NOT** skip the measurement step -- every experiment must be measured.\n9. **DO NOT** keep changes that regress the metric unless the user explicitly allowed trade-offs.\n10. **DO NOT** install new dependencies or make environment changes unless the user approved it.\n\n---\n\n## Phase 1: Setup (Interactive)\n\nBefore any experimentation begins, work with the user to establish these parameters.\nAsk the user directly for each item. Do not assume or skip any.\n\n### 1.1 Define the Goal\n\nAsk the user:\n\n> **What are you trying to improve or optimize?**\n>\n> Examples: execution time, memory usage, binary size, test pass rate, code coverage,\n> API response latency, throughput, error rate, benchmark score, build time, bundle size,\n> lines of code, cyclomatic complexity, etc.\n\nRecord the user's answer as the **goal**.\n\n### 1.2 Define the Metric\n\nAsk the user:\n\n> **How do we measure success? What exact command produces the metric?**\n>\n> I need:\n> 1. **The command** to run (e.g., `dotnet test`, `npm run benchmark`, `time .\u002Fbuild.sh`, `pytest --tb=short`)\n> 2. **How to extract the metric** from the output (e.g., a regex pattern, a specific line, a JSON field)\n> 3. **Direction**: Is lower better or higher better?\n>\n> Example: \"Run `dotnet test --logger trx`, count passing tests. Higher is better.\"\n> Example: \"Run `hyperfine '.\u002Fmy-program'`, extract mean time. Lower is better.\"\n\nRecord:\n- `METRIC_COMMAND`: the command to run\n- `METRIC_EXTRACTION`: how to extract the numeric metric from output\n- `METRIC_DIRECTION`: `lower_is_better` or `higher_is_better`\n\n### 1.3 Define the Scope\n\nAsk the user:\n\n> **Which files or directories am I allowed to modify?**\n>\n> And which files are OFF LIMITS (read-only)?\n\nRecord:\n- `IN_SCOPE_FILES`: files\u002Fdirs the agent may edit\n- `OUT_OF_SCOPE_FILES`: files\u002Fdirs that must not be modified\n\n### 1.4 Define Constraints\n\nAsk the user:\n\n> **Are there any constraints I should respect?**\n>\n> Examples:\n> - Time budget per experiment (e.g., \"each run should take \u003C 2 minutes\")\n> - No new dependencies\n> - Must keep all existing tests passing\n> - Must not change the public API\n> - Must maintain backward compatibility\n> - VRAM\u002Fmemory limit\n> - Code complexity limits (prefer simpler solutions)\n\nRecord as `CONSTRAINTS`.\n\n### 1.5 Define the Experiment Budget (Optional)\n\nAsk the user:\n\n> **How many experiments should I run, or should I just keep going until you stop me?**\n>\n> You can say a number (e.g., \"try 20 experiments\") or \"unlimited\" (I'll run until you interrupt).\n\nRecord as `MAX_EXPERIMENTS` (number or `unlimited`).\n\n### 1.6 Simplicity Criterion\n\nInform the user of the default simplicity policy:\n\n> **Simplicity policy (default):** All else being equal, simpler is better. A small improvement\n> that adds ugly complexity is not worth it. Removing code while maintaining or improving\n> the metric is a great outcome. I'll weigh the complexity cost against the improvement\n> magnitude. Does this policy work for you, or do you want to adjust it?\n\nRecord any adjustments as `SIMPLICITY_POLICY`.\n\n### 1.7 Confirm Setup\n\nSummarize all parameters back to the user in a clear table:\n\n| Parameter          | Value                        |\n| ------------------ | ---------------------------- |\n| Goal               | ...                          |\n| Metric command     | ...                          |\n| Metric extraction  | ...                          |\n| Direction          | lower is better \u002F higher ... |\n| In-scope files     | ...                          |\n| Out-of-scope files | ...                          |\n| Constraints        | ...                          |\n| Max experiments    | ...                          |\n| Simplicity policy  | ...                          |\n\nAsk the user to confirm. Do not proceed until confirmed.\n\n---\n\n## Phase 2: Branch & Baseline\n\nOnce the user confirms:\n\n1. **Create a branch**: Propose a tag based on today's date (e.g., `autoresearch\u002Fmar17`).\n   Create the branch: `git checkout -b autoresearch\u002F\u003Ctag>`.\n\n2. **Read in-scope files**: Read all files that are in scope to build full context of the current state.\n\n3. **Initialize results.tsv**: Create `results.tsv` in the repo root with the header row:\n   ```\n   experiment\tcommit\tmetric\tstatus\tdescription\n   ```\n   Add `results.tsv` and `run.log` to `.git\u002Finfo\u002Fexclude` (append if not already present) so they stay untracked without modifying any tracked files.\n\n4. **Run the baseline**: Execute the metric command on the current unmodified code.\n   Record the result as experiment `0` with status `baseline` in `results.tsv`.\n\n5. **Report baseline** to the user:\n   > Baseline established: **[metric_name] = [value]**\n   > Starting autonomous experimentation loop.\n\n---\n\n## Phase 3: Experiment Loop\n\nRun this loop continuously. Do not stop to ask the user. Run until:\n- `MAX_EXPERIMENTS` is reached, OR\n- The user manually interrupts\n\n### For each experiment:\n\n```\nLOOP:\n  1. THINK   - Analyze previous results and the current code.\n               Generate an experiment hypothesis.\n               Consider: what worked, what didn't, what hasn't been tried.\n\n  2. EDIT    - Modify the in-scope file(s) to implement the idea.\n               Keep changes focused and minimal per experiment.\n\n  3. COMMIT  - git add + git commit with a short descriptive message.\n               Format: \"experiment: \u003Cshort description of what changed>\"\n\n  4. RUN     - Execute the metric command.\n               Redirect output to run.log so it does not flood the context window.\n               Use shell-appropriate redirection:\n               - Bash\u002FZsh: `\u003Ccommand> > run.log 2>&1`\n               - PowerShell: `\u003Ccommand> *> run.log`\n\n  5. MEASURE - Extract the metric from run.log.\n               If extraction fails (crash\u002Ferror), read the last 50 lines\n               of run.log for the error.\n\n  6. DECIDE  - Compare metric to the current best:\n               - IMPROVED: Keep the commit. Update the \"best\" baseline.\n                 Log status = \"keep\".\n               - SAME OR WORSE: Revert. `git reset --hard HEAD~1`.\n                 Log status = \"discard\".\n               - CRASH: Attempt a quick fix (typo, import, simple error).\n                 Amend the experiment commit (`git commit --amend`) with the fix\n                 and rerun. The experiment keeps its original number.\n                 If unfixable after 2 attempts, revert the entire experiment\n                 (`git reset --hard HEAD~1`) and log status = \"crash\".\n\n  7. LOG     - Append a row to results.tsv:\n               experiment_number  commit_hash  metric_value  status  description\n\n  8. CONTINUE - Go to step 1.\n```\n\n### Experiment Strategy\n\nWhen generating experiment ideas, follow this priority order:\n\n1. **Low-hanging fruit first**: Simple parameter tweaks, obvious inefficiencies.\n2. **Informed by results**: If a direction showed promise, explore further in that direction.\n3. **Diversify after plateaus**: If the last 3-5 experiments all failed, try a different approach entirely.\n4. **Combine winners**: If experiments A and B each improved independently, try combining them.\n5. **Simplification passes**: Periodically try removing code\u002Fcomplexity to see if the metric holds.\n6. **Radical changes**: After exhausting incremental ideas, try larger architectural changes.\n\n### Handling Constraints\n\n- **Time budget**: If a run exceeds 2x the expected duration, kill it and treat as a crash.\n- **Existing tests**: If constraints require tests to pass, run them before\u002Fafter and revert if they break.\n- **Memory\u002Fresources**: Monitor and revert if resource usage exceeds stated limits.\n\n---\n\n## Phase 4: Reporting\n\nWhen the loop ends (budget reached or user interrupts):\n\n1. **Print the full results.tsv** as a formatted table.\n2. **Summarize**:\n   - Total experiments run\n   - Experiments kept \u002F discarded \u002F crashed\n   - Starting metric (baseline) vs. final metric\n   - Improvement percentage\n   - Top 3 most impactful changes\n3. **Show the cumulative git log** of kept experiments:\n   `git log --oneline \u003Cstart_commit>..HEAD`\n4. **Recommend next steps**: Based on the results, suggest what a human researcher might try next (ideas that were too risky\u002Fcomplex for automated experimentation).\n\n---\n\n## Quick Reference\n\n### Results TSV Format\n\nTab-separated, 5 columns:\n\n```\nexperiment\tcommit\tmetric\tstatus\tdescription\n0\ta1b2c3d\t0.997900\tbaseline\tunmodified code\n1\tb2c3d4e\t0.993200\tkeep\tincrease learning rate to 0.04\n2\tc3d4e5f\t1.005000\tdiscard\tswitch to GeLU activation\n3\td4e5f6g\t0.000000\tcrash\tdouble model width (OOM)\n```\n\n### Git Workflow\n\n- All experiments happen on the `autoresearch\u002F\u003Ctag>` branch\n- Each experiment is committed before running\n- Failed experiments are reverted with `git reset --hard HEAD~1`\n- Successful experiments advance the branch\n- `results.tsv` and `run.log` stay untracked (added to `.git\u002Finfo\u002Fexclude`)\n\n### Key Principles\n\n1. **Measure everything**: No experiment without a measurement.\n2. **Revert failures**: The branch only advances on improvements.\n3. **Stay autonomous**: Never stop to ask. Think harder if stuck.\n4. **Keep it simple**: Complexity is a cost. Weigh it against gains.\n5. **Log everything**: The TSV is the research journal.\n",{"name":5,"license":15,"metadata":61,"description":7,"compatibility":64},{"author":62,"inspired-by":63},"luiscantero","https:\u002F\u002Fgithub.com\u002Fkarpathy\u002Fautoresearch","Requires git. The project must be a git repository. Requires terminal access to run commands.","main","5269f7ea2d6c3a9bfda5ed878b6726cf509121fd","skills\u002Fautoresearch",5344,"https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot","https:\u002F\u002Fawesome-copilot.github.com\u002F","https:\u002F\u002Fgithub.com\u002Fgithub","https:\u002F\u002Fgithub.com\u002Fabout","How people build software.","2026-11-13 02:05:36",{"source":8,"skillName":5,"commands":76,"prompt":86,"zipAvailable":87},[77,80,83],{"tool":78,"command":79},"npx","npx skills add github\u002Fawesome-copilot --skill autoresearch",{"tool":81,"command":82},"bunx","bunx skills add github\u002Fawesome-copilot --skill autoresearch",{"tool":84,"command":85},"pnpm","pnpm dlx skills add github\u002Fawesome-copilot --skill autoresearch","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fautoresearch\u002FSKILL.md",true,[89,95,101,105,109,113],{"path":90,"resourceType":91,"resourceTypeDesc":92,"sizeBytes":93,"sourceHtmlUrl":94},"assertions.md","REFERENCE","参考文档",4226,"https:\u002F\u002Fgithub.com\u002Fdavila7\u002Fclaude-code-templates\u002Fblob\u002Fmain\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fdevelopment\u002Fplaywright-java\u002Fassertions.md",{"path":96,"resourceType":97,"resourceTypeDesc":98,"sizeBytes":99,"sourceHtmlUrl":100},"BasePage.java","SCRIPT","脚本",4830,"https:\u002F\u002Fgithub.com\u002Fdavila7\u002Fclaude-code-templates\u002Fblob\u002Fmain\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fdevelopment\u002Fplaywright-java\u002FBasePage.java",{"path":102,"resourceType":97,"resourceTypeDesc":98,"sizeBytes":103,"sourceHtmlUrl":104},"BaseTest.java",4526,"https:\u002F\u002Fgithub.com\u002Fdavila7\u002Fclaude-code-templates\u002Fblob\u002Fmain\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fdevelopment\u002Fplaywright-java\u002FBaseTest.java",{"path":106,"resourceType":91,"resourceTypeDesc":92,"sizeBytes":107,"sourceHtmlUrl":108},"config.md",6651,"https:\u002F\u002Fgithub.com\u002Fdavila7\u002Fclaude-code-templates\u002Fblob\u002Fmain\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fdevelopment\u002Fplaywright-java\u002Fconfig.md",{"path":110,"resourceType":91,"resourceTypeDesc":92,"sizeBytes":111,"sourceHtmlUrl":112},"fixtures.md",7755,"https:\u002F\u002Fgithub.com\u002Fdavila7\u002Fclaude-code-templates\u002Fblob\u002Fmain\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fdevelopment\u002Fplaywright-java\u002Ffixtures.md",{"path":114,"resourceType":91,"resourceTypeDesc":92,"sizeBytes":115,"sourceHtmlUrl":116},"page-objects.md",6104,"https:\u002F\u002Fgithub.com\u002Fdavila7\u002Fclaude-code-templates\u002Fblob\u002Fmain\u002Fcli-tool\u002Fcomponents\u002Fskills\u002Fdevelopment\u002Fplaywright-java\u002Fpage-objects.md",[118,147,175,200,233,249],{"skillNo":119,"name":120,"nameZh":6,"description":121,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":122,"categories":123,"keywords":128,"publishedAt":56},"01M70KXZNX9ES38N2A8M58ZZMZ","supabase-rls-policies","Write, review, and test Supabase Row Level Security (RLS) policies so that one user cannot read or change another user's rows. Use when creating or changing tables, policies, views, or security definer functions in a Supabase project, when reviewing a migration that touches RLS, or when asked to prove user isolation with pgTAP tests that sign in as two users.","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fsupabase-rls-policies\u002FSKILL.md",[124,127],{"categoryNo":125,"categoryName":126,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000008","测试与质量",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[129,130,131,132,35,133,54,134,135,136,137,138,139,140,141,142,143,144,145,146],"supabase","rls","supabase rls","supabase rls policies","when","change","read","write","security","isolation","tables","as","tests","prove","use","users","functions","with",{"skillNo":148,"name":149,"nameZh":6,"description":150,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":151,"categories":152,"keywords":154,"publishedAt":56},"01M70KWA5YWXDV6V8DZ19E3T2Y","skill-distribution-audit","Audit a skill collection before release or investigate installer omissions, archived\u002Ftemplate leakage, duplicate identities, and manifest drift. Reconcile the intended public source set, distribution manifest, and actual discovery output for a specified consumer version. Includes a deterministic set-comparison helper.","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fskill-distribution-audit\u002FSKILL.md",[153],{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[155,156,157,158,159,160,161,162,163,164,165,166,167,168,169,170,171,172,173,174],"distribution","skill distribution audit","audit","SET","deterministic","installer","for","discovery","comparison","helper","manifest","version","collection","drift","output","investigate","duplicate","template","release","reconcile",{"skillNo":176,"name":177,"nameZh":6,"description":178,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":179,"sourceHtmlUrl":180,"categories":181,"keywords":183,"publishedAt":56},"01M70KCJPFVW7PPQ49CC5R8JGK","logicapps-custom-code-net10-migration","Migrate Azure Logic Apps Standard local-function custom-code projects from .NET 8 to .NET 10. Use for legacy Logic Apps custom-code workspace upgrades, project\u002Fpackage configuration, custom-worker settings, publication, and integrated VS Code debugging. Not for ordinary .NET applications, standalone Azure Functions, Consumption workflows, or cloud deployment.","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Flogicapps-custom-code-net10-migration\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Flogicapps-custom-code-net10-migration\u002FSKILL.md",[182],{"categoryNo":27,"categoryName":28,"taxonomyType":29,"taxonomyTypeDesc":30},[184,47,185,186,187,188,161,35,189,190,191,192,193,194,195,143,196,197,198,199],"custom","logicapps custom code net10 migration","migration","projects","standalone","deployment","applications","local","azure","worker","upgrades","package","apps","Legacy","debugging","workflows",{"skillNo":201,"name":202,"nameZh":6,"description":203,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":204,"sourceHtmlUrl":205,"categories":206,"keywords":211,"publishedAt":56},"01M6AEBN37RQMXJ3298PEY7PG1","winmd-api-search","Find and explore Windows desktop APIs. Use when building features that need platform capabilities — camera, file access, notifications, UI controls, AI\u002FML, sensors, networking, etc. Discovers the right API for a task and retrieves full type details (methods, properties, events, enumeration values).","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fwinmd-api-search\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fwinmd-api-search\u002FSKILL.md",[207,210],{"categoryNo":208,"categoryName":209,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000010","后端与接口",{"categoryNo":27,"categoryName":28,"taxonomyType":29,"taxonomyTypeDesc":30},[212,213,214,215,216,217,218,219,220,221,222,223,224,225,41,133,226,227,228,229,230,161,231,232],"winmd api search","search","api","windows","desktop","file access","enumeration","ui","ml","file","networking","capabilities","notifications","explore","features","full","camera","events","platform","methods","controls",{"skillNo":234,"name":235,"nameZh":6,"description":236,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":237,"licenseFileUrl":238,"sourceHtmlUrl":239,"categories":240,"keywords":246,"publishedAt":56},"01M6AE571R3SC1ZGE202B3NB5S","resemble-detect","Deepfake detection and media safety — detect AI-generated audio, images, video, and text, trace synthesis sources, apply watermarks, verify speaker identity, and analyze media intelligence using Resemble AI","Apache-2.0","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fresemble-detect\u002FLICENSE","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fresemble-detect\u002FSKILL.md",[241,242,245],{"categoryNo":208,"categoryName":209,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":243,"categoryName":244,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000016","图像与多媒体",{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[247,158,221,32,248],"resemble detect","identity",{"skillNo":250,"name":251,"nameZh":6,"description":252,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":237,"licenseFileUrl":253,"sourceHtmlUrl":254,"categories":255,"keywords":258,"publishedAt":56},"01M6AE30BSK2HCY4TQQ22Z8JRD","quality-playbook","Run a complete quality engineering audit on any codebase. Derives behavioral requirements from the code, generates spec-traced functional tests, runs a three-pass code review with regression tests, executes a multi-model spec audit (Council of Three), and produces a consolidated bug report with TDD-verified patches. Finds the 35% of real defects that structural code review alone cannot catch. Works with any language. Trigger on 'quality playbook', 'spec audit', 'Council of Three', 'fitness-to-purpose', or 'coverage theater'.","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fquality-playbook\u002FLICENSE.txt","https:\u002F\u002Fgithub.com\u002Fgithub\u002Fawesome-copilot\u002Fblob\u002Fmain\u002Fskills\u002Fquality-playbook\u002FSKILL.md",[256,257],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},{"categoryNo":24,"categoryName":25,"taxonomyType":21,"taxonomyTypeDesc":22},[259,260,261,44,47,54,9,157,262,48,32,263,264,265,266,267,268,269,37,270,271,272,141,273,274,275,276],"playbook","quality playbook","quality","tdd","report","spec","bug report","council","engineering","functional","spec audit","verified","multi-model","three","coverage","model","regression","structural"]