[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7C3MBTRCSW0KR92WYYVQMDA":3,"skill-same:01M7C3MBTRCSW0KR92WYYVQMDA":45},{"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":24,"commitSha":25,"skillDirectory":26,"forksCount":27,"repositoryHtmlUrl":28,"repositoryHomepageUrl":6,"ownerHtmlUrl":29,"ownerBlogUrl":6,"ownerDescription":6,"updatedAt":30,"install":31,"resources":44},"01M7C3MBTRCSW0KR92WYYVQMDA","skilltta-task-guidance",null,"Task-specific operational guidance synthesized at test time from retrieved training examples.","linyouxue\u002Fskill-repair-benchmark","linyouxue","https:\u002F\u002Favatars.githubusercontent.com\u002Fu\u002F216151464?v=4","01M7BJEVCXKD74KH6FG9CBVSQB","林宥学","01M7BJEVE02NP217V93BKZFC0J",0,"Apache-2.0","https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark\u002Fblob\u002Fmain\u002Fevaluation\u002Fresults\u002Fskilltta-synthesis-opus47-tatumass-failed15-20260923\u002Ftasks\u002Fmanufacturing-equipment-maintenance\u002Frepaired_skill\u002Fskilltta-task-guidance\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: skilltta-task-guidance\ndescription: Task-specific operational guidance synthesized at test time from retrieved training examples.\n---\n\n# SKILL.md\n\n## When to use\nUse this skill when a task requires answering multi-part manufacturing\u002Fprocess-engineering questions grounded in a domain **handbook (PDF)** plus supporting **structured data files** (e.g., CSV logs, sensor traces, defect tables), and requires writing JSON answer files to a specific output directory in a strict schema.\n\nTarget-contract essentials (at a high level):\n- All thresholds, definitions, equations, temperature regions, time windows, geometry, dwell requirements, and selection rules must come from the handbook — never guessed or assumed from prior knowledge.\n- Numeric answers must be computed from the provided data files (sensor, MES, defects), not fabricated.\n- Each sub-question produces its own JSON file with a strictly specified schema, float rounding, sort order, and `null` policy.\n- \"Best per group\" or ranking questions require you to combine handbook-defined quality\u002Fcompliance criteria with efficiency signals from the data.\n\nEvidence to gather before acting:\n1. Extract handbook text (PDF → text) and search for the exact section names referenced in each question (e.g., preheat definition, ramp equation, liquidus, TAL window, peak temperature spec, conveyor\u002Fdwell rules, tie-breakers, thermocouple selection rule).\n2. Inspect every CSV: columns, units, run_id keys, timestamps or time axis, thermocouple ids, any join keys (run_id ↔ board_family ↔ board_id).\n3. Verify units before computing (°C vs K, s vs ms, cm\u002Fmin vs mm\u002Fs, sample rate for ramp = dT\u002Fdt).\n4. Identify the handbook's rule for choosing a \"representative\" thermocouple when several are present (e.g., component vs board, hottest, coldest, specific channel).\n\n## Possible Failure Modes\n- Hard-coding thresholds (ramp limit, liquidus temperature, min\u002Fmax TAL, min peak, conveyor speed bounds, board dimensions) from memory instead of extracting them from the handbook.\n- Using the wrong temperature window for \"preheat\" — computing ramp across the whole profile rather than only the handbook-defined preheat region.\n- Computing ramp rate as (Tmax − Tmin)\u002FΔt over the region instead of the handbook's specified equation (often a sliding\u002Finstantaneous dT\u002Fdt, sometimes over a fixed window).\n- Selecting the wrong thermocouple per run (e.g., averaging all TCs, or picking max, when the handbook prescribes a specific rule — often \"worst-case\" or a named channel).\n- TAL computation errors: not interpolating crossings of the liquidus line, double-counting multiple excursions incorrectly, or ignoring runs with no data (should be flagged, not silently skipped).\n- Peak temperature: treating missing thermocouple data as \"pass\" or as null instead of the required \"failing\" per the contract.\n- Conveyor feasibility: mixing units (cm vs mm, min vs s), or ignoring that required dwell time comes from summing handbook zone requirements, then converting via board length.\n- \"Best run\" selection: inventing a scoring formula rather than following handbook's stated quality\u002Fefficiency priorities and tie-breakers; forgetting that non-compliant runs are typically ineligible.\n- Output schema drift: missing keys, wrong nesting, unsorted arrays, unrounded floats, using `\"None\"` string instead of JSON `null`, or omitting `runner_up_run_ids` when empty.\n- Writing outputs to the wrong directory or filename; not creating `\u002Fapp\u002Foutput\u002F`.\n- Copying values, ids, or structural conventions from unrelated retrieved examples — the retrieved examples here are generic Python coding tasks and contribute no domain facts.\n\n## Possible procedures\n1. **Ingest the handbook once, cache the facts.** Use a PDF text extractor; grep for keywords per question (preheat, ramp, liquidus, TAL, peak, conveyor, dwell, zone, board, tie-breaker). Record each numeric threshold with its unit and section reference in a small dict you reuse across questions.\n2. **Load CSVs with pandas.** Confirm dtypes and time column. Group thermocouple data by `run_id` and `tc_id`. Confirm every run in MES has (or does not have) thermocouple coverage — this drives null\u002Ffailing logic.\n3. **Per-run temperature analysis pipeline:**\n   - Slice the profile to the handbook-defined region (preheat \u002F above-liquidus \u002F peak).\n   - Apply the handbook's exact equation (e.g., ramp = ΔT\u002FΔt on consecutive samples; TAL = time where T ≥ liquidus, interpolated at boundary crossings).\n   - Apply the handbook's TC-selection rule to pick the representative `tc_id`.\n4. **Compliance decisions** must reference handbook thresholds only; do not invent tolerance. Encode limits as named constants sourced from the handbook.\n5. **Conveyor feasibility:** required_min (or max) speed = board_length \u002F required_total_dwell (watch units). Compare to actual speed from MES.\n6. **Best-per-family selection:** first filter to compliant runs across all prior checks; then rank by the handbook's stated primary metric (often defect rate or yield from `test_defects.csv`), break ties by efficiency (speed\u002Fthroughput). Every other run in the family (compliant or not, unless spec says otherwise) goes into `runner_up_run_ids`, sorted ascending.\n7. **Write JSON:** build dicts matching the exact schema shown in the prompt; round floats with `round(x, 2)`; sort lists by the specified key; ensure `null` where data is absent; `os.makedirs('\u002Fapp\u002Foutput', exist_ok=True)` before writing.\n8. **Recovery checks:**\n   - If a handbook value cannot be located, re-search with alternative keywords\u002Fsynonyms before defaulting; do not guess.\n   - If a run has zero thermocouple rows, follow the contract's \"no data ⇒ failing\" rule and still emit the run in output with appropriate nulls\u002FFalse.\n   - Sanity-check ramp signs, TAL non-negativity, and that peak ≥ liquidus for any compliant run.\n\n## Verification Checklist\n- [ ] Every threshold, equation, region, window, and selection rule used is quoted\u002Ftraceable to the handbook (not assumed).\n- [ ] All five output files exist at `\u002Fapp\u002Foutput\u002Fq0{1..5}.json` with the exact top-level structure specified in the prompt.\n- [ ] Floats rounded to 2 decimals; arrays sorted ascending by `run_id` \u002F `board_family` \u002F `board_id` as specified.\n- [ ] Missing data uses JSON `null`, not strings; booleans are real `true`\u002F`false`.\n- [ ] Runs without thermocouple data are handled per the contract (e.g., failing for peak check) and still appear in outputs.\n- [ ] Units verified (°C\u002Fs for ramps, seconds for TAL, cm\u002Fmin for conveyor); no unit-mixing.\n- [ ] Representative thermocouple per run chosen by handbook rule, not by ad-hoc heuristic.\n- [ ] Best-run selection filters by compliance first, then applies handbook's ranking + tie-break; `runner_up_run_ids` present (possibly empty list) and sorted.\n- [ ] No values, ids, or schema fragments copied from retrieved examples — they are unrelated.\n- [ ] JSON files parse (`json.load` round-trip) and contain every required key even when values are null\u002Fempty.\n- [ ] Unrelated files under `\u002Fapp\u002Fdata\u002F` are not modified; only `\u002Fapp\u002Foutput\u002F` is written.\n- [ ] If any handbook fact was unresolved, re-check the PDF before falling back — never fabricate a limit.\n","# SKILL.md\n\n## When to use\nUse this skill when a task requires answering multi-part manufacturing\u002Fprocess-engineering questions grounded in a domain **handbook (PDF)** plus supporting **structured data files** (e.g., CSV logs, sensor traces, defect tables), and requires writing JSON answer files to a specific output directory in a strict schema.\n\nTarget-contract essentials (at a high level):\n- All thresholds, definitions, equations, temperature regions, time windows, geometry, dwell requirements, and selection rules must come from the handbook — never guessed or assumed from prior knowledge.\n- Numeric answers must be computed from the provided data files (sensor, MES, defects), not fabricated.\n- Each sub-question produces its own JSON file with a strictly specified schema, float rounding, sort order, and `null` policy.\n- \"Best per group\" or ranking questions require you to combine handbook-defined quality\u002Fcompliance criteria with efficiency signals from the data.\n\nEvidence to gather before acting:\n1. Extract handbook text (PDF → text) and search for the exact section names referenced in each question (e.g., preheat definition, ramp equation, liquidus, TAL window, peak temperature spec, conveyor\u002Fdwell rules, tie-breakers, thermocouple selection rule).\n2. Inspect every CSV: columns, units, run_id keys, timestamps or time axis, thermocouple ids, any join keys (run_id ↔ board_family ↔ board_id).\n3. Verify units before computing (°C vs K, s vs ms, cm\u002Fmin vs mm\u002Fs, sample rate for ramp = dT\u002Fdt).\n4. Identify the handbook's rule for choosing a \"representative\" thermocouple when several are present (e.g., component vs board, hottest, coldest, specific channel).\n\n## Possible Failure Modes\n- Hard-coding thresholds (ramp limit, liquidus temperature, min\u002Fmax TAL, min peak, conveyor speed bounds, board dimensions) from memory instead of extracting them from the handbook.\n- Using the wrong temperature window for \"preheat\" — computing ramp across the whole profile rather than only the handbook-defined preheat region.\n- Computing ramp rate as (Tmax − Tmin)\u002FΔt over the region instead of the handbook's specified equation (often a sliding\u002Finstantaneous dT\u002Fdt, sometimes over a fixed window).\n- Selecting the wrong thermocouple per run (e.g., averaging all TCs, or picking max, when the handbook prescribes a specific rule — often \"worst-case\" or a named channel).\n- TAL computation errors: not interpolating crossings of the liquidus line, double-counting multiple excursions incorrectly, or ignoring runs with no data (should be flagged, not silently skipped).\n- Peak temperature: treating missing thermocouple data as \"pass\" or as null instead of the required \"failing\" per the contract.\n- Conveyor feasibility: mixing units (cm vs mm, min vs s), or ignoring that required dwell time comes from summing handbook zone requirements, then converting via board length.\n- \"Best run\" selection: inventing a scoring formula rather than following handbook's stated quality\u002Fefficiency priorities and tie-breakers; forgetting that non-compliant runs are typically ineligible.\n- Output schema drift: missing keys, wrong nesting, unsorted arrays, unrounded floats, using `\"None\"` string instead of JSON `null`, or omitting `runner_up_run_ids` when empty.\n- Writing outputs to the wrong directory or filename; not creating `\u002Fapp\u002Foutput\u002F`.\n- Copying values, ids, or structural conventions from unrelated retrieved examples — the retrieved examples here are generic Python coding tasks and contribute no domain facts.\n\n## Possible procedures\n1. **Ingest the handbook once, cache the facts.** Use a PDF text extractor; grep for keywords per question (preheat, ramp, liquidus, TAL, peak, conveyor, dwell, zone, board, tie-breaker). Record each numeric threshold with its unit and section reference in a small dict you reuse across questions.\n2. **Load CSVs with pandas.** Confirm dtypes and time column. Group thermocouple data by `run_id` and `tc_id`. Confirm every run in MES has (or does not have) thermocouple coverage — this drives null\u002Ffailing logic.\n3. **Per-run temperature analysis pipeline:**\n   - Slice the profile to the handbook-defined region (preheat \u002F above-liquidus \u002F peak).\n   - Apply the handbook's exact equation (e.g., ramp = ΔT\u002FΔt on consecutive samples; TAL = time where T ≥ liquidus, interpolated at boundary crossings).\n   - Apply the handbook's TC-selection rule to pick the representative `tc_id`.\n4. **Compliance decisions** must reference handbook thresholds only; do not invent tolerance. Encode limits as named constants sourced from the handbook.\n5. **Conveyor feasibility:** required_min (or max) speed = board_length \u002F required_total_dwell (watch units). Compare to actual speed from MES.\n6. **Best-per-family selection:** first filter to compliant runs across all prior checks; then rank by the handbook's stated primary metric (often defect rate or yield from `test_defects.csv`), break ties by efficiency (speed\u002Fthroughput). Every other run in the family (compliant or not, unless spec says otherwise) goes into `runner_up_run_ids`, sorted ascending.\n7. **Write JSON:** build dicts matching the exact schema shown in the prompt; round floats with `round(x, 2)`; sort lists by the specified key; ensure `null` where data is absent; `os.makedirs('\u002Fapp\u002Foutput', exist_ok=True)` before writing.\n8. **Recovery checks:**\n   - If a handbook value cannot be located, re-search with alternative keywords\u002Fsynonyms before defaulting; do not guess.\n   - If a run has zero thermocouple rows, follow the contract's \"no data ⇒ failing\" rule and still emit the run in output with appropriate nulls\u002FFalse.\n   - Sanity-check ramp signs, TAL non-negativity, and that peak ≥ liquidus for any compliant run.\n\n## Verification Checklist\n- [ ] Every threshold, equation, region, window, and selection rule used is quoted\u002Ftraceable to the handbook (not assumed).\n- [ ] All five output files exist at `\u002Fapp\u002Foutput\u002Fq0{1..5}.json` with the exact top-level structure specified in the prompt.\n- [ ] Floats rounded to 2 decimals; arrays sorted ascending by `run_id` \u002F `board_family` \u002F `board_id` as specified.\n- [ ] Missing data uses JSON `null`, not strings; booleans are real `true`\u002F`false`.\n- [ ] Runs without thermocouple data are handled per the contract (e.g., failing for peak check) and still appear in outputs.\n- [ ] Units verified (°C\u002Fs for ramps, seconds for TAL, cm\u002Fmin for conveyor); no unit-mixing.\n- [ ] Representative thermocouple per run chosen by handbook rule, not by ad-hoc heuristic.\n- [ ] Best-run selection filters by compliance first, then applies handbook's ranking + tie-break; `runner_up_run_ids` present (possibly empty list) and sorted.\n- [ ] No values, ids, or schema fragments copied from retrieved examples — they are unrelated.\n- [ ] JSON files parse (`json.load` round-trip) and contain every required key even when values are null\u002Fempty.\n- [ ] Unrelated files under `\u002Fapp\u002Fdata\u002F` are not modified; only `\u002Fapp\u002Foutput\u002F` is written.\n- [ ] If any handbook fact was unresolved, re-check the PDF before falling back — never fabricate a limit.\n",{"name":5,"description":7},"main","cdcb6d230f3b15c3c33f8a2bdd16439f43602cbe","evaluation\u002Fresults\u002Fskilltta-synthesis-opus47-tatumass-failed15-20260923\u002Ftasks\u002Fmanufacturing-equipment-maintenance\u002Frepaired_skill\u002Fskilltta-task-guidance",1,"https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark","https:\u002F\u002Fgithub.com\u002Flinyouxue","2026-11-14 20:28:35",{"source":8,"skillName":5,"commands":32,"prompt":42,"zipAvailable":43},[33,36,39],{"tool":34,"command":35},"npx","npx skills add linyouxue\u002Fskill-repair-benchmark --skill skilltta-task-guidance",{"tool":37,"command":38},"bunx","bunx skills add linyouxue\u002Fskill-repair-benchmark --skill skilltta-task-guidance",{"tool":40,"command":41},"pnpm","pnpm dlx skills add linyouxue\u002Fskill-repair-benchmark --skill skilltta-task-guidance","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark\u002Fblob\u002Fmain\u002Fevaluation\u002Fresults\u002Fskilltta-synthesis-opus47-tatumass-failed15-20260923\u002Ftasks\u002Fmanufacturing-equipment-maintenance\u002Frepaired_skill\u002Fskilltta-task-guidance\u002FSKILL.md",true,[],[46,53,60,67,72,77],{"skillNo":47,"name":48,"nameZh":6,"description":49,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":50,"categories":51,"keywords":52,"publishedAt":19},"01M7E1HGVRY73NBJ022CJG13X6","video-frame-extraction","Extract frames from video files and save them as images using OpenCV","https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark\u002Fblob\u002Fmain\u002Fjpg-ocr-stat\u002Fresult\u002Fskilltta-claude4.7\u002Fgroup_03\u002Frun_03\u002Finputs\u002Fskills\u002Fvideo-frame-extraction\u002FSKILL.md",[],[],{"skillNo":54,"name":55,"nameZh":6,"description":56,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":57,"categories":58,"keywords":59,"publishedAt":19},"01M7E1HAG5R7CJ9C8NYJS771MB","openai-vision","Analyze images and multi-frame sequences using OpenAI GPT vision models","https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark\u002Fblob\u002Fmain\u002Fjpg-ocr-stat\u002Fresult\u002Fskilltta-claude4.7\u002Fgroup_03\u002Frun_03\u002Finputs\u002Fskills\u002Fopenai-vision\u002FSKILL.md",[],[],{"skillNo":61,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":64,"categories":65,"keywords":66,"publishedAt":19},"01M7E1H7DADKB6QBWKC2KQ63SZ","image-ocr","Extract text content from images using Tesseract OCR via Python, including structured field extraction from receipts and forms","https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark\u002Fblob\u002Fmain\u002Fjpg-ocr-stat\u002Fresult\u002Fskilltta-claude4.7\u002Fgroup_03\u002Frun_03\u002Finputs\u002Fskills\u002Fimage-ocr\u002FSKILL.md",[],[],{"skillNo":68,"name":48,"nameZh":6,"description":49,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":69,"categories":70,"keywords":71,"publishedAt":19},"01M7E1H0WBNT43TENGQY7AVZHD","https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark\u002Fblob\u002Fmain\u002Fjpg-ocr-stat\u002Fresult\u002Fskilltta-claude4.7\u002Fgroup_03\u002Frun_02\u002Finputs\u002Fskills\u002Fvideo-frame-extraction\u002FSKILL.md",[],[],{"skillNo":73,"name":55,"nameZh":6,"description":56,"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},"01M7E1GTC7D4N4Q2EF03875J9E","https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark\u002Fblob\u002Fmain\u002Fjpg-ocr-stat\u002Fresult\u002Fskilltta-claude4.7\u002Fgroup_03\u002Frun_02\u002Finputs\u002Fskills\u002Fopenai-vision\u002FSKILL.md",[],[],{"skillNo":78,"name":62,"nameZh":6,"description":63,"descriptionZh":6,"repositoryFullName":8,"ownerLogin":9,"ownerAvatarUrl":10,"ownerAccountNo":11,"ownerDisplayName":12,"repositoryNo":13,"starsCount":14,"licenseSpdxId":15,"licenseFileUrl":6,"sourceHtmlUrl":79,"categories":80,"keywords":81,"publishedAt":19},"01M7E1GQ92CWFMGH53KSSE4WER","https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark\u002Fblob\u002Fmain\u002Fjpg-ocr-stat\u002Fresult\u002Fskilltta-claude4.7\u002Fgroup_03\u002Frun_02\u002Finputs\u002Fskills\u002Fimage-ocr\u002FSKILL.md",[],[]]