[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M6HV86J8CSM5MENQ85RY3FAX":3,"skill-same:01M6HV86J8CSM5MENQ85RY3FAX":71},{"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":23,"publishedAt":44,"linkOnly":45,"markdownContent":46,"markdownBody":47,"frontMatter":48,"refName":51,"commitSha":52,"skillDirectory":53,"forksCount":54,"repositoryHtmlUrl":55,"repositoryHomepageUrl":6,"ownerHtmlUrl":56,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":57,"install":58,"resources":70},"01M6HV86J8CSM5MENQ85RY3FAX","prompt-optimizer",null,"Optimize LLM prompt templates against an eval harness. Works with autoresearch for overnight prompt optimization. For projects that call LLMs (try-on prompts, chatProxy, MCP tools).","Silex-Research\u002FDontPanic","Silex-Research","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F130925891?v=4","01M6HV5K48NKSZT37SQKSV639D","Silex Research","01M6HV5K8R2281GSJB6S9A5TF8",5,"Apache-2.0","https:\u002F\u002Fgithub.com\u002FSilex-Research\u002FDontPanic\u002Fblob\u002Fmain\u002Fclaude\u002Fskills\u002Fprompt-optimizer\u002FSKILL.md",[18],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000023","Skill与智能体开发","SCENARIO","使用场景",[24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43],"optimizer","prompt optimizer","prompt","eval","for","true","Prompt Optimization","pattern","call","optimize","command","MCP","projects","Tools","Optimization","autoresearch","file","harness","prompt templates","llms","2026-11-10 10:12:30",false,"---\nname: prompt-optimizer\ndescription: Optimize LLM prompt templates against an eval harness. Works with autoresearch for overnight prompt optimization. For projects that call LLMs (try-on prompts, chatProxy, MCP tools).\ndisable-model-invocation: true\nargument-hint: \u003Cprompt_file> --eval \u003Ceval_command> [--metric \u003Cpattern>] [--iterations \u003Cn>]\n---\n\n# Prompt Optimizer — Iterative Prompt Improvement\n\nYou are a prompt engineer. Your job is to systematically improve LLM prompt templates by testing variations against an evaluation harness.\n\n## Inputs (from $ARGUMENTS)\n\n| Param | Default | Description |\n|-------|---------|-------------|\n| prompt_file | required | Path to the prompt template file |\n| --eval | required | Command to evaluate prompt quality |\n| --metric | \"score\" | Grep pattern to extract the metric from eval output |\n| --iterations | 10 | Number of variations to try |\n| --direction | higher | Metric direction: `higher` or `lower` |\n\n## Protocol\n\n### 1. Baseline\n- Read the current prompt template\n- Run the eval command to establish baseline metric\n- Log: `baseline | \u003Cmetric_value> | \u003Cprompt_hash>`\n\n### 2. Analyze\n- Identify potential improvements:\n  - Clarity: ambiguous instructions, missing context\n  - Specificity: vague criteria, missing examples\n  - Structure: ordering of instructions, formatting\n  - Constraints: missing guardrails, edge cases\n  - Few-shot examples: missing, weak, or misleading\n\n### 3. Generate Variations\nFor each iteration:\n- Apply ONE change at a time (isolate variables)\n- Categories of changes:\n  - **Reorder** — move instructions to different positions\n  - **Rephrase** — say the same thing differently\n  - **Add constraint** — add a guardrail or boundary\n  - **Add example** — add a few-shot example\n  - **Remove noise** — delete redundant or confusing instructions\n  - **Simplify** — reduce complexity without losing intent\n\n### 4. Evaluate\n- Run the eval command with the modified prompt\n- Extract the metric\n- Compare to best-so-far\n\n### 5. Keep or Discard\n- If metric improved: keep the change, update best-so-far\n- If metric worsened or unchanged: discard the change\n\n### 6. Report\n```\nPROMPT OPTIMIZATION REPORT — \u003Cprompt_file>\n==========================================\nBaseline: \u003Cinitial_metric>\nBest:     \u003Cfinal_metric> (+\u003Cimprovement>%)\nIterations: \u003Cn>\nKept changes: \u003Ck>\n\nChanges applied:\n  1. [+0.05] Reordered composition instructions before style instructions\n  2. [+0.03] Added explicit \"do not crop\" constraint\n  3. [+0.02] Simplified person identity section\n\nChanges rejected:\n  1. [-0.02] Added few-shot example (confused the model)\n  2. [+0.00] Rephrased quality instructions (no effect)\n```\n\n## Integration with Autoresearch\n\nThis skill can be used as the \"strategy\" layer inside an autoresearch loop:\n```\n\u002Fautoresearch \"\u003Ceval_command>\" --target-files \"\u003Cprompt_file>\" --metric \"\u003Cpattern>\" --direction higher\n```\n\nWhen used standalone, it runs a fixed number of iterations. When paired with autoresearch, autoresearch handles the git branching and iteration loop while this skill guides what changes to try.\n\n## Rules\n\n- Change ONE thing per iteration — never stack multiple changes\n- Always run the eval, never assume a change is an improvement\n- Keep a copy of the original prompt — don't lose the starting point\n- Log every variation attempted, not just the ones that worked\n- Prompt optimization has diminishing returns — stop when 3 consecutive iterations show no improvement\n","# Prompt Optimizer — Iterative Prompt Improvement\n\nYou are a prompt engineer. Your job is to systematically improve LLM prompt templates by testing variations against an evaluation harness.\n\n## Inputs (from $ARGUMENTS)\n\n| Param | Default | Description |\n|-------|---------|-------------|\n| prompt_file | required | Path to the prompt template file |\n| --eval | required | Command to evaluate prompt quality |\n| --metric | \"score\" | Grep pattern to extract the metric from eval output |\n| --iterations | 10 | Number of variations to try |\n| --direction | higher | Metric direction: `higher` or `lower` |\n\n## Protocol\n\n### 1. Baseline\n- Read the current prompt template\n- Run the eval command to establish baseline metric\n- Log: `baseline | \u003Cmetric_value> | \u003Cprompt_hash>`\n\n### 2. Analyze\n- Identify potential improvements:\n  - Clarity: ambiguous instructions, missing context\n  - Specificity: vague criteria, missing examples\n  - Structure: ordering of instructions, formatting\n  - Constraints: missing guardrails, edge cases\n  - Few-shot examples: missing, weak, or misleading\n\n### 3. Generate Variations\nFor each iteration:\n- Apply ONE change at a time (isolate variables)\n- Categories of changes:\n  - **Reorder** — move instructions to different positions\n  - **Rephrase** — say the same thing differently\n  - **Add constraint** — add a guardrail or boundary\n  - **Add example** — add a few-shot example\n  - **Remove noise** — delete redundant or confusing instructions\n  - **Simplify** — reduce complexity without losing intent\n\n### 4. Evaluate\n- Run the eval command with the modified prompt\n- Extract the metric\n- Compare to best-so-far\n\n### 5. Keep or Discard\n- If metric improved: keep the change, update best-so-far\n- If metric worsened or unchanged: discard the change\n\n### 6. Report\n```\nPROMPT OPTIMIZATION REPORT — \u003Cprompt_file>\n==========================================\nBaseline: \u003Cinitial_metric>\nBest:     \u003Cfinal_metric> (+\u003Cimprovement>%)\nIterations: \u003Cn>\nKept changes: \u003Ck>\n\nChanges applied:\n  1. [+0.05] Reordered composition instructions before style instructions\n  2. [+0.03] Added explicit \"do not crop\" constraint\n  3. [+0.02] Simplified person identity section\n\nChanges rejected:\n  1. [-0.02] Added few-shot example (confused the model)\n  2. [+0.00] Rephrased quality instructions (no effect)\n```\n\n## Integration with Autoresearch\n\nThis skill can be used as the \"strategy\" layer inside an autoresearch loop:\n```\n\u002Fautoresearch \"\u003Ceval_command>\" --target-files \"\u003Cprompt_file>\" --metric \"\u003Cpattern>\" --direction higher\n```\n\nWhen used standalone, it runs a fixed number of iterations. When paired with autoresearch, autoresearch handles the git branching and iteration loop while this skill guides what changes to try.\n\n## Rules\n\n- Change ONE thing per iteration — never stack multiple changes\n- Always run the eval, never assume a change is an improvement\n- Keep a copy of the original prompt — don't lose the starting point\n- Log every variation attempted, not just the ones that worked\n- Prompt optimization has diminishing returns — stop when 3 consecutive iterations show no improvement\n",{"name":5,"description":7,"argument-hint":49,"disable-model-invocation":50},"\u003Cprompt_file> --eval \u003Ceval_command> [--metric \u003Cpattern>] [--iterations \u003Cn>]",true,"main","14ffaf54d57448d72d9f32efca397086f3de0e64","claude\u002Fskills\u002Fprompt-optimizer",1,"https:\u002F\u002Fgithub.com\u002FSilex-Research\u002FDontPanic","https:\u002F\u002Fgithub.com\u002FSilex-Research","2026-11-10 08:27:00",{"source":8,"skillName":5,"commands":59,"prompt":69,"zipAvailable":50},[60,63,66],{"tool":61,"command":62},"npx","npx skills add Silex-Research\u002FDontPanic --skill prompt-optimizer",{"tool":64,"command":65},"bunx","bunx skills add Silex-Research\u002FDontPanic --skill prompt-optimizer",{"tool":67,"command":68},"pnpm","pnpm dlx skills add Silex-Research\u002FDontPanic --skill prompt-optimizer","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002FSilex-Research\u002FDontPanic\u002Fblob\u002Fmain\u002Fclaude\u002Fskills\u002Fprompt-optimizer\u002FSKILL.md",[],[72,87,108,124,142,160],{"skillNo":73,"name":74,"nameZh":6,"description":75,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":76,"categories":77,"keywords":78,"publishedAt":44},"01M6HVB95QJ3TXWYQ4M6F1SGCJ","worktree-isolation","Creates an isolated git worktree.","https:\u002F\u002Fgithub.com\u002FSilex-Research\u002FDontPanic\u002Fblob\u002Fmain\u002Fscripts\u002Fdontpanic_orchestrate\u002Ftests\u002Ffixtures\u002Fskill_rubrics\u002Fworktree-isolation\u002FSKILL.md",[],[79,80,81,82,83,84,85,86],"worktree isolation","isolation","worktree","repo","mutation","implementation","git worktree","git",{"skillNo":88,"name":89,"nameZh":6,"description":90,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":91,"categories":92,"keywords":96,"publishedAt":44},"01M6HVB49HTWVCD3DBH0Y1D84E","test-runner","Runs the targeted test suite.","https:\u002F\u002Fgithub.com\u002FSilex-Research\u002FDontPanic\u002Fblob\u002Fmain\u002Fscripts\u002Fdontpanic_orchestrate\u002Ftests\u002Ffixtures\u002Fskill_rubrics\u002Ftest-runner\u002FSKILL.md",[93],{"categoryNo":94,"categoryName":95,"taxonomyType":21,"taxonomyTypeDesc":22},"01JBSEEDCAT000000000000008","测试与质量",[97,98,99,100,101,102,103,104,105,84,106,107],"test","test runner","runner","review","safe","auto","path","evidence","test suite","pytest","output",{"skillNo":109,"name":110,"nameZh":6,"description":111,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":112,"categories":113,"keywords":114,"publishedAt":44},"01M6HVAZMH62Z9GTM8GRDN2MEA","subagent-driven-dev","Meta toolkit for dispatching subagents.","https:\u002F\u002Fgithub.com\u002FSilex-Research\u002FDontPanic\u002Fblob\u002Fmain\u002Fscripts\u002Fdontpanic_orchestrate\u002Ftests\u002Ffixtures\u002Fskill_rubrics\u002Fsubagent-driven-dev\u002FSKILL.md",[],[115,116,117,118,119,120,121,28,122,123],"dev","subagent","driven","subagent driven dev","subagents","dispatching subagents","never","Meta","toolkit",{"skillNo":125,"name":126,"nameZh":6,"description":127,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":128,"categories":129,"keywords":130,"publishedAt":44},"01M6HVAWTYK3JF6Q7PGBTQBBZH","security-review","Read-only security review of pending changes.","https:\u002F\u002Fgithub.com\u002FSilex-Research\u002FDontPanic\u002Fblob\u002Fmain\u002Fscripts\u002Fdontpanic_orchestrate\u002Ftests\u002Ffixtures\u002Fskill_rubrics\u002Fsecurity-review\u002FSKILL.md",[],[131,132,100,133,134,126,135,136,137,102,138,139,104,140,141],"security review","security","files","run","json","read","skills run","readonly","only","read-only","changes",{"skillNo":143,"name":144,"nameZh":6,"description":145,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":146,"categories":147,"keywords":148,"publishedAt":44},"01M6HVARJ3DCMZH3RZFHAEFJJJ","revenue-check","Reads production revenue dashboards.","https:\u002F\u002Fgithub.com\u002FSilex-Research\u002FDontPanic\u002Fblob\u002Fmain\u002Fscripts\u002Fdontpanic_orchestrate\u002Ftests\u002Ffixtures\u002Fskill_rubrics\u002Frevenue-check\u002FSKILL.md",[],[149,150,151,152,100,153,154,155,156,157,135,136,158,104,159],"revenue","revenue check","check","project","planning","network","Production","dashboards","access","approval","id",{"skillNo":161,"name":5,"nameZh":6,"description":162,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":163,"categories":164,"keywords":166,"publishedAt":44},"01M6HVANBY28H9E38P1T01EEQ5","Optimizes a prompt template.","https:\u002F\u002Fgithub.com\u002FSilex-Research\u002FDontPanic\u002Fblob\u002Fmain\u002Fscripts\u002Fdontpanic_orchestrate\u002Ftests\u002Ffixtures\u002Fskill_rubrics\u002Fprompt-optimizer\u002FSKILL.md",[165],{"categoryNo":19,"categoryName":20,"taxonomyType":21,"taxonomyTypeDesc":22},[25,26,24,167,168,84],"template","text"]