[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7C33CCMYZRWE01XT4KA1S54":3,"skill-same:01M7C33CCMYZRWE01XT4KA1S54":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},"01M7C33CCMYZRWE01XT4KA1S54","manufacturing-failure-reason-codebook-normalization",null,"This skill should be considered when you need to normalize testing engineers' written defect reasons following the provided product codebooks. This skill will correct the typos, misused abbreviations, ambiguous descriptions, mixed Chinese-English text or misleading text and provide explanations. This skill will do segmentation, semantic matching, confidence calibration and station validation.","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\u002FWML\u002Ftasks\u002Fmanufacturing-codebook-normalization\u002Frepaired_skill\u002Fmanufacturing-failure-reason-codebook-normalization\u002FSKILL.md",[],[],"2026-11-15 14:19:17",false,"---\nname: manufacturing-failure-reason-codebook-normalization\ndescription: This skill should be considered when you need to normalize testing engineers' written defect reasons following the provided product codebooks. This skill will correct the typos, misused abbreviations, ambiguous descriptions, mixed Chinese-English text or misleading text and provide explanations. This skill will do segmentation, semantic matching, confidence calibration and station validation.\n---\n\nThis skill should be considered when you need to normalize, standardize, or correct testing engineers' written failure reasons to match the requirements provided in the product codebooks. Common errors in engineer-written reasons include ambiguous descriptions, missing important words, improper personal writing habits, using wrong abbreviations, improper combining multiple reasons into one sentence without clear spacing or in wrong order, writing wrong station names or model, writing typos, improper combining Chinese and English characters, cross-project differences, and taking wrong products' codebook.\n\nSome codes are defined for specific stations and cannot be used by other stations. If entry.stations is not None, the predicted code should only be considered valid when the record station matches one of the stations listed in entry.stations. Otherwise, the code should be rejected. For each record segment, the system evaluates candidate codes defined in the corresponding product codebook and computes an internal matching score for each candidate. You should consider multiple evidence sources to calculate the score to measure how well a candidate code explains the segment, and normalize the score to a stable range [0.0, 1.0]. Evidence can include text evidence from raw_reason_text (e.g., overlap or fuzzy similarity between span_text and codebook text such as standard_label, keywords_examples, or categories), station compatibility, fail_code alignment, test_item alignment, and conflict cues such as mutually exclusive or contradictory signals.\n\nEvaluator-aligned text evidence accounting (for non-UNKNOWN decisions): When using any augmentation\u002Fnormalization or synonym bridging for scoring, treat it as supportive only. Before emitting any non-UNKNOWN code, compute a simple lexical-support proxy that matches the delivered output constraints: build a token set from the raw span_text (exact substring) and a token set from the predicted code’s codebook fields (standard_label + keywords_examples + categories), and check that the intersection is non-trivial.\n\nLexical-support gate (mandatory): for any non-UNKNOWN decision, require minimal raw-token overlap under the same tokenization as the delivered text.\n- Compute Jaccard overlap: ov = |tokens(span_text) ∩ tokens(code_fields)| \u002F |tokens(span_text) ∪ tokens(code_fields)|.\n- If ov \u003C 0.01, do NOT emit that code (regardless of semantic\u002Ffuzzy score). This is a hard final gate.\n- If ov >= 0.01, the code is eligible; higher ov should generally permit higher confidence.\n\nNotes\u002Fnegative boundary:\n- Do not “count” tokens introduced only by augmentation unless the same tokens also appear in span_text.\n- Do not rewrite span_text to force overlap; span_text must remain an exact substring.\n- This gate applies only to deciding whether to emit a non-UNKNOWN code; UNKNOWN segments are allowed to have ov=0.\n\nAfter all candidate codes are scored, sort them in descending order. Let c1 be the top candidate with score s1 and c2 be the second candidate with score s2.\n\nDeterministic tie-break: When multiple candidates fall within a small margin of the best score, the system applies a deterministic tie-break based on record context (e.g., record_id, segment index, station, fail_code, test_item) to avoid always choosing the same code in near-tie cases while keeping outputs reproducible. Apply tie-break only after the top candidate clears an absolute evidence floor (both a reasonable match score and the lexical-support proxy described above). If all near-best candidates are weakly supported, skip tie-break and output UNKNOWN.\n\nTo provide convincing answers, add station, fail_code, test_item, a short token overlap cue, or a component reference to the rationale.\n\nUNKNOWN handling: UNKNOWN should be decided based on the best match only (i.e., after ranking), not by marking multiple candidates. If the best-match score is low (weak evidence), output pred_code=\"UNKNOWN\" and pred_label=\"\" to give engineering an alert. Additionally, even when the best-match score is not extremely low, do not emit a known code unless there is minimal lexical support in the raw span_text for that predicted codebook entry (token overlap with the codebook’s label\u002Fkeywords\u002Fcategories). This ensures non-UNKNOWN outputs remain grounded in the segment’s literal text.\n\nFallback\u002Fverification step: If a segment’s top candidate fails the lexical-support gate, follow this deterministic ladder (station-compatible candidates only):\n1) Walk down the ranked list (c2, c3, ...) and find the first candidate that passes the lexical-support gate (ov >= 0.01).\n2) Emit that candidate only if its score is not materially worse than the top candidate (i.e., within a reasonable score gap you already use for near-ties); otherwise output UNKNOWN.\n3) If no candidate passes ov >= 0.01, output UNKNOWN.\n\nKeep span_text unchanged as an exact substring.\n\nWhen strong positive cues exist (e.g., clear component references like R\u002FC\u002FU\u002FL designators, explicit CRC\u002FChecksum wording, explicit fixture\u002Fpogo\u002Fprobe\u002Fcontact wording that appears in span_text), UNKNOWN should be less frequent than in generic or noisy segments.\n\nConfidence calibration: confidence ranges from 0.0 to 1.0 and reflects an engineering confidence level (not a probability). Calibrate confidence from match quality so that UNKNOWN predictions are generally less confident than non-UNKNOWN predictions, and confidence values are not nearly constant. Confidence should show distribution-level separation between UNKNOWN and non-UNKNOWN predictions (e.g., means, quantiles, and diversity), and should be weakly aligned with evidence strength; round confidence to 4 decimals.\n\nHere is a pipeline reference\n1) Load test_center_logs.csv into logs_rows and load each product codebook; build valid_code_set, station_scope_map, and CodebookEntry objects.  \n2) For each record, split raw_reason_text into 1–N segments; each segment uses segment_id=\u003Crecord_id>-S\u003Ci> and keeps an exact substring as span_text.  \n3) For each segment, filter candidates by station scope, then compute match score from combined evidence (text evidence, station compatibility, context alignment, and conflict cues).  \n4) Rank candidates by score; apply the lexical-support gate (ov >= 0.01) as a final eligibility check for any non-UNKNOWN; if the top candidate fails, fall back to the next-ranked eligible candidate within a reasonable score gap; otherwise output UNKNOWN. Apply deterministic tie-break only among eligible candidates that meet the absolute evidence floor.  \n5) Output exactly one pred_code\u002Fpred_label per segment from the product codebook (or UNKNOWN\u002F\"\" when best evidence is weak or lacks minimal raw-text lexical support) and compute confidence by calibrating match quality with sufficient diversity; round to 4 decimals.\n\nRepresentative run result:\n{\"task_name\": \"manufacturing-codebook-normalization\", \"rollout_name\": \"manufacturing-codebook-normalization__20d53cd3\", \"rewards\": {\"reward\": 0.0}, \"agent\": \"openhands\", \"agent_name\": \"OpenHands CLI ACP Agent\", \"model\": \"openrouter\u002Fopenai\u002Fgpt-5.2\", \"skill_mode\": \"with-skill\", \"skill_source\": \"task_bundled\", \"requested_skills_dir\": \"\u002Fhome\u002Flinyuanjing\u002Fskillsbench-v1.1\u002Ftasks\u002Fmanufacturing-codebook-normalization\u002Fenvironment\u002Fskills\", \"effective_skills_dir\": \"\u002Ftmp\u002Fbenchflow-task-e_hh299x\u002Fmanufacturing-codebook-normalization\u002Fenvironment\u002Fskills\", \"skills_sandbox_dir\": \"\u002Fskills\", \"include_task_skills\": true, \"n_tool_calls\": 32, \"n_skill_invocations\": 1, \"n_prompts\": 3, \"agent_result\": {\"n_tool_calls\": 32, \"n_skill_invocations\": 1, \"n_prompts\": 3, \"n_input_tokens\": 0, \"n_output_tokens\": 0, \"n_cache_read_tokens\": 0, \"n_cache_creation_tokens\": 0, \"total_tokens\": 1288292, \"cost_usd\": 0.67804275, \"usage_source\": \"provider_response\", \"price_source\": \"litellm\"}, \"final_metrics\": {\"total_prompt_tokens\": 0, \"total_completion_tokens\": 0, \"total_cached_tokens\": 0, \"total_cost_usd\": 0.67804275}, \"trajectory_summary\": {\"steps\": 41, \"tool_call_steps\": 32, \"user_message_steps\": 3, \"agent_message_steps\": 3, \"agent_thought_steps\": 3, \"other_steps\": 0, \"event_type_counts\": {\"user_message\": 3, \"agent_thought\": 3, \"tool_call\": 32, \"agent_message\": 3}, \"tool_call_status_counts\": {\"completed\": 32}, \"partial_trajectory\": false, \"trajectory_source\": \"acp\"}, \"usage_tracking\": {\"requested\": \"required\", \"status\": \"enabled\", \"environment\": \"docker\", \"endpoint_kind\": \"host\", \"usage_source\": \"provider_response\"}, \"error\": null, \"error_category\": null, \"verifier_error\": null, \"verifier_error_category\": null, \"export_error\": null, \"idle_timeout_info\": null, \"agent_timeout_info\": null, \"sandbox_startup_info\": null, \"transport_error_info\": null, \"verifier_timeout_info\": null, \"api_error_info\": null, \"suspected_api_error_info\": null, \"partial_trajectory\": false, \"trajectory_source\": \"acp\", \"started_at\": \"2026-08-23 10:39:42.436130\", \"finished_at\": \"2026-08-23 10:53:59.051112\", \"timing\": {\"environment_setup\": 74.2, \"agent_setup\": 28.0, \"agent_execution\": 483.0, \"verifier\": 30.3, \"total\": 856.6}, \"scenes\": [{\"name\": \"default\", \"skills_dir\": null, \"roles\": [{\"name\": \"agent\", \"agent\": \"openhands\", \"model\": \"openrouter\u002Fopenai\u002Fgpt-5.2\", \"reasoning_effort\": null, \"timeout_sec\": null, \"idle_timeout_sec\": null, \"skills_dir\": null, \"capabilities\": null, \"env_keys\": []}], \"turns\": [{\"role\": \"agent\", \"has_prompt\": true}, {\"role\": \"agent\", \"has_prompt\": true}, {\"role\": \"agent\", \"has_prompt\": true}]}], \"loop\": {\"strategy\": \"single-shot\"}, \"task_digest\": \"sha256:07f8f87ed53b8224fefc65dc220537572980e0794970db30e3873aebd063c9d3\", \"sandbox_id\": null}\n","This skill should be considered when you need to normalize, standardize, or correct testing engineers' written failure reasons to match the requirements provided in the product codebooks. Common errors in engineer-written reasons include ambiguous descriptions, missing important words, improper personal writing habits, using wrong abbreviations, improper combining multiple reasons into one sentence without clear spacing or in wrong order, writing wrong station names or model, writing typos, improper combining Chinese and English characters, cross-project differences, and taking wrong products' codebook.\n\nSome codes are defined for specific stations and cannot be used by other stations. If entry.stations is not None, the predicted code should only be considered valid when the record station matches one of the stations listed in entry.stations. Otherwise, the code should be rejected. For each record segment, the system evaluates candidate codes defined in the corresponding product codebook and computes an internal matching score for each candidate. You should consider multiple evidence sources to calculate the score to measure how well a candidate code explains the segment, and normalize the score to a stable range [0.0, 1.0]. Evidence can include text evidence from raw_reason_text (e.g., overlap or fuzzy similarity between span_text and codebook text such as standard_label, keywords_examples, or categories), station compatibility, fail_code alignment, test_item alignment, and conflict cues such as mutually exclusive or contradictory signals.\n\nEvaluator-aligned text evidence accounting (for non-UNKNOWN decisions): When using any augmentation\u002Fnormalization or synonym bridging for scoring, treat it as supportive only. Before emitting any non-UNKNOWN code, compute a simple lexical-support proxy that matches the delivered output constraints: build a token set from the raw span_text (exact substring) and a token set from the predicted code’s codebook fields (standard_label + keywords_examples + categories), and check that the intersection is non-trivial.\n\nLexical-support gate (mandatory): for any non-UNKNOWN decision, require minimal raw-token overlap under the same tokenization as the delivered text.\n- Compute Jaccard overlap: ov = |tokens(span_text) ∩ tokens(code_fields)| \u002F |tokens(span_text) ∪ tokens(code_fields)|.\n- If ov \u003C 0.01, do NOT emit that code (regardless of semantic\u002Ffuzzy score). This is a hard final gate.\n- If ov >= 0.01, the code is eligible; higher ov should generally permit higher confidence.\n\nNotes\u002Fnegative boundary:\n- Do not “count” tokens introduced only by augmentation unless the same tokens also appear in span_text.\n- Do not rewrite span_text to force overlap; span_text must remain an exact substring.\n- This gate applies only to deciding whether to emit a non-UNKNOWN code; UNKNOWN segments are allowed to have ov=0.\n\nAfter all candidate codes are scored, sort them in descending order. Let c1 be the top candidate with score s1 and c2 be the second candidate with score s2.\n\nDeterministic tie-break: When multiple candidates fall within a small margin of the best score, the system applies a deterministic tie-break based on record context (e.g., record_id, segment index, station, fail_code, test_item) to avoid always choosing the same code in near-tie cases while keeping outputs reproducible. Apply tie-break only after the top candidate clears an absolute evidence floor (both a reasonable match score and the lexical-support proxy described above). If all near-best candidates are weakly supported, skip tie-break and output UNKNOWN.\n\nTo provide convincing answers, add station, fail_code, test_item, a short token overlap cue, or a component reference to the rationale.\n\nUNKNOWN handling: UNKNOWN should be decided based on the best match only (i.e., after ranking), not by marking multiple candidates. If the best-match score is low (weak evidence), output pred_code=\"UNKNOWN\" and pred_label=\"\" to give engineering an alert. Additionally, even when the best-match score is not extremely low, do not emit a known code unless there is minimal lexical support in the raw span_text for that predicted codebook entry (token overlap with the codebook’s label\u002Fkeywords\u002Fcategories). This ensures non-UNKNOWN outputs remain grounded in the segment’s literal text.\n\nFallback\u002Fverification step: If a segment’s top candidate fails the lexical-support gate, follow this deterministic ladder (station-compatible candidates only):\n1) Walk down the ranked list (c2, c3, ...) and find the first candidate that passes the lexical-support gate (ov >= 0.01).\n2) Emit that candidate only if its score is not materially worse than the top candidate (i.e., within a reasonable score gap you already use for near-ties); otherwise output UNKNOWN.\n3) If no candidate passes ov >= 0.01, output UNKNOWN.\n\nKeep span_text unchanged as an exact substring.\n\nWhen strong positive cues exist (e.g., clear component references like R\u002FC\u002FU\u002FL designators, explicit CRC\u002FChecksum wording, explicit fixture\u002Fpogo\u002Fprobe\u002Fcontact wording that appears in span_text), UNKNOWN should be less frequent than in generic or noisy segments.\n\nConfidence calibration: confidence ranges from 0.0 to 1.0 and reflects an engineering confidence level (not a probability). Calibrate confidence from match quality so that UNKNOWN predictions are generally less confident than non-UNKNOWN predictions, and confidence values are not nearly constant. Confidence should show distribution-level separation between UNKNOWN and non-UNKNOWN predictions (e.g., means, quantiles, and diversity), and should be weakly aligned with evidence strength; round confidence to 4 decimals.\n\nHere is a pipeline reference\n1) Load test_center_logs.csv into logs_rows and load each product codebook; build valid_code_set, station_scope_map, and CodebookEntry objects.  \n2) For each record, split raw_reason_text into 1–N segments; each segment uses segment_id=\u003Crecord_id>-S\u003Ci> and keeps an exact substring as span_text.  \n3) For each segment, filter candidates by station scope, then compute match score from combined evidence (text evidence, station compatibility, context alignment, and conflict cues).  \n4) Rank candidates by score; apply the lexical-support gate (ov >= 0.01) as a final eligibility check for any non-UNKNOWN; if the top candidate fails, fall back to the next-ranked eligible candidate within a reasonable score gap; otherwise output UNKNOWN. Apply deterministic tie-break only among eligible candidates that meet the absolute evidence floor.  \n5) Output exactly one pred_code\u002Fpred_label per segment from the product codebook (or UNKNOWN\u002F\"\" when best evidence is weak or lacks minimal raw-text lexical support) and compute confidence by calibrating match quality with sufficient diversity; round to 4 decimals.\n\nRepresentative run result:\n{\"task_name\": \"manufacturing-codebook-normalization\", \"rollout_name\": \"manufacturing-codebook-normalization__20d53cd3\", \"rewards\": {\"reward\": 0.0}, \"agent\": \"openhands\", \"agent_name\": \"OpenHands CLI ACP Agent\", \"model\": \"openrouter\u002Fopenai\u002Fgpt-5.2\", \"skill_mode\": \"with-skill\", \"skill_source\": \"task_bundled\", \"requested_skills_dir\": \"\u002Fhome\u002Flinyuanjing\u002Fskillsbench-v1.1\u002Ftasks\u002Fmanufacturing-codebook-normalization\u002Fenvironment\u002Fskills\", \"effective_skills_dir\": \"\u002Ftmp\u002Fbenchflow-task-e_hh299x\u002Fmanufacturing-codebook-normalization\u002Fenvironment\u002Fskills\", \"skills_sandbox_dir\": \"\u002Fskills\", \"include_task_skills\": true, \"n_tool_calls\": 32, \"n_skill_invocations\": 1, \"n_prompts\": 3, \"agent_result\": {\"n_tool_calls\": 32, \"n_skill_invocations\": 1, \"n_prompts\": 3, \"n_input_tokens\": 0, \"n_output_tokens\": 0, \"n_cache_read_tokens\": 0, \"n_cache_creation_tokens\": 0, \"total_tokens\": 1288292, \"cost_usd\": 0.67804275, \"usage_source\": \"provider_response\", \"price_source\": \"litellm\"}, \"final_metrics\": {\"total_prompt_tokens\": 0, \"total_completion_tokens\": 0, \"total_cached_tokens\": 0, \"total_cost_usd\": 0.67804275}, \"trajectory_summary\": {\"steps\": 41, \"tool_call_steps\": 32, \"user_message_steps\": 3, \"agent_message_steps\": 3, \"agent_thought_steps\": 3, \"other_steps\": 0, \"event_type_counts\": {\"user_message\": 3, \"agent_thought\": 3, \"tool_call\": 32, \"agent_message\": 3}, \"tool_call_status_counts\": {\"completed\": 32}, \"partial_trajectory\": false, \"trajectory_source\": \"acp\"}, \"usage_tracking\": {\"requested\": \"required\", \"status\": \"enabled\", \"environment\": \"docker\", \"endpoint_kind\": \"host\", \"usage_source\": \"provider_response\"}, \"error\": null, \"error_category\": null, \"verifier_error\": null, \"verifier_error_category\": null, \"export_error\": null, \"idle_timeout_info\": null, \"agent_timeout_info\": null, \"sandbox_startup_info\": null, \"transport_error_info\": null, \"verifier_timeout_info\": null, \"api_error_info\": null, \"suspected_api_error_info\": null, \"partial_trajectory\": false, \"trajectory_source\": \"acp\", \"started_at\": \"2026-08-23 10:39:42.436130\", \"finished_at\": \"2026-08-23 10:53:59.051112\", \"timing\": {\"environment_setup\": 74.2, \"agent_setup\": 28.0, \"agent_execution\": 483.0, \"verifier\": 30.3, \"total\": 856.6}, \"scenes\": [{\"name\": \"default\", \"skills_dir\": null, \"roles\": [{\"name\": \"agent\", \"agent\": \"openhands\", \"model\": \"openrouter\u002Fopenai\u002Fgpt-5.2\", \"reasoning_effort\": null, \"timeout_sec\": null, \"idle_timeout_sec\": null, \"skills_dir\": null, \"capabilities\": null, \"env_keys\": []}], \"turns\": [{\"role\": \"agent\", \"has_prompt\": true}, {\"role\": \"agent\", \"has_prompt\": true}, {\"role\": \"agent\", \"has_prompt\": true}]}], \"loop\": {\"strategy\": \"single-shot\"}, \"task_digest\": \"sha256:07f8f87ed53b8224fefc65dc220537572980e0794970db30e3873aebd063c9d3\", \"sandbox_id\": null}\n",{"name":5,"description":7},"main","cdcb6d230f3b15c3c33f8a2bdd16439f43602cbe","evaluation\u002Fresults\u002FWML\u002Ftasks\u002Fmanufacturing-codebook-normalization\u002Frepaired_skill\u002Fmanufacturing-failure-reason-codebook-normalization",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 manufacturing-failure-reason-codebook-normalization",{"tool":37,"command":38},"bunx","bunx skills add linyouxue\u002Fskill-repair-benchmark --skill manufacturing-failure-reason-codebook-normalization",{"tool":40,"command":41},"pnpm","pnpm dlx skills add linyouxue\u002Fskill-repair-benchmark --skill manufacturing-failure-reason-codebook-normalization","安装技能，技能地址：https:\u002F\u002Fgithub.com\u002Flinyouxue\u002Fskill-repair-benchmark\u002Fblob\u002Fmain\u002Fevaluation\u002Fresults\u002FWML\u002Ftasks\u002Fmanufacturing-codebook-normalization\u002Frepaired_skill\u002Fmanufacturing-failure-reason-codebook-normalization\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",[],[]]