[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"skill:01M7C3XBXKNMJ9K7PH12CS74HC":3,"skill-same:01M7C3XBXKNMJ9K7PH12CS74HC":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},"01M7C3XBXKNMJ9K7PH12CS74HC","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\u002Fvideo-silence-remover\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 the target task provides a media file (video\u002Faudio) and asks you to detect and remove low-information segments (silence, static openings, long pauses), then produce both a processed media artifact and a structured JSON report describing what was removed. The binding contract for this class of task typically requires:\n- A processed output file at a specified workspace path.\n- A JSON report with a fixed schema documenting original\u002Fcompressed\u002Fremoved durations, a compression percentage, and a list of removed segments with start\u002Fend\u002Fduration.\n- Internal arithmetic consistency (original ≈ compressed + removed).\n- Preservation of substantive content while trimming an opening and long pauses above some threshold.\n\nBefore acting, gather:\n- Exact input path, exact output filenames, and the exact JSON schema keys\u002Ftypes from the task prompt.\n- The pause-length threshold and any qualitative guidance about the opening (e.g., static frames, noise).\n- Time\u002Fcompute limits mentioned (e.g., must finish under a stated wall-clock budget).\n- Available tools in the environment (ffmpeg, ffprobe, Python audio libs).\n\n## Possible Failure Modes\n- Schema drift: renaming keys, nesting differently, or omitting required fields like `compression_percentage` or per-segment `duration`.\n- Inconsistent arithmetic: `original_duration_seconds` not ≈ `compressed_duration_seconds + removed_duration_seconds`, or `compression_percentage` computed against the wrong base.\n- Overly aggressive silence detection that cuts speech pauses shorter than the stated threshold, removing teaching content.\n- Overly conservative detection that leaves the static opening or long pauses intact, yielding too-low compression.\n- Detecting silence only by video (dark\u002Fstatic frames) or only by audio when the opening is characterized by both static video and noisy audio — missing one signal.\n- Reporting segments in the compressed timeline rather than in original-video time coordinates.\n- Merging adjacent silent segments incorrectly (or not at all), producing many tiny cuts and audible glitches.\n- Re-encoding with a slow codec \u002F high quality settings that blow the runtime budget, or using stream copy across cut points and getting keyframe-misaligned output.\n- Off-by-one\u002Fprecision issues: writing integer seconds when floats are expected, or float noise causing the sum check to fail.\n- Writing outputs to the wrong directory (not the current workspace) or with slightly different filenames than required.\n- Leaving temp files or partially written JSON if the pipeline errors midway.\n\n## Possible procedures\n1. Probe the input:\n   - Use `ffprobe` to get exact duration, streams, sample rate, and codec.\n   - Extract audio (mono, e.g., 16 kHz WAV) for analysis; keep the original for final cutting.\n\n2. Detect the opening:\n   - Use audio energy\u002FRMS over a sliding window and\u002For `ffmpeg -af silencedetect` from t=0 to find the first sustained speech onset.\n   - Optionally corroborate with a video signal (frame-difference or scene score) to confirm static frames at the start.\n   - Trim from 0 up to shortly before the first confident speech onset; avoid cutting into the first spoken words (leave a small pad, e.g., 0.2–0.5s).\n\n3. Detect long pauses:\n   - Run silence detection with a threshold and minimum-duration parameter matched to the task’s pause criterion (e.g., > the stated seconds).\n   - Sensible starting points: noise floor around -30 to -35 dB, min silence duration equal to (or slightly above) the task threshold.\n   - Merge overlapping\u002Fadjacent silence intervals; add a small keep-pad on each side so speech isn’t clipped.\n\n4. Build the removal plan in original-time coordinates:\n   - Produce a list of `[start, end]` segments to remove (opening + long pauses), sorted and non-overlapping.\n   - Compute keep segments as the complement within `[0, original_duration]`.\n\n5. Cut and concatenate:\n   - Prefer ffmpeg with a single filter graph (`select`\u002F`atrempo`-free `aselect` + `setpts`\u002F`asetpts` and `concat`) or generate segment files and concat-demux them.\n   - Re-encode with a fast but reasonable codec (e.g., libx264 `-preset veryfast`, AAC audio) to guarantee clean cuts and stay within the time budget.\n   - Verify the output file plays and has expected duration via `ffprobe`.\n\n6. Emit the JSON report:\n   - Use exactly the keys and nesting shown in the prompt.\n   - Populate durations from measured values: `original_duration_seconds` from ffprobe of input; `compressed_duration_seconds` from ffprobe of output; `removed_duration_seconds` = sum of removed segment durations; `compression_percentage` = removed \u002F original * 100 (confirm which base the prompt implies).\n   - Ensure each removed segment includes `start`, `end`, `duration` in original-video seconds, sorted ascending, non-overlapping.\n   - Round consistently (e.g., to 2–3 decimals) so the arithmetic check still holds.\n\n7. Sanity checks and recovery:\n   - If compression is implausibly high (e.g., > ~50%) or low (~0%), re-tune thresholds and re-run detection rather than shipping.\n   - If sum check fails, recompute from the segment list rather than patching numbers.\n   - If runtime approaches the budget, switch to faster ffmpeg presets or coarser analysis windows.\n\n## Verification Checklist\n- [ ] Both required output artifacts exist at the exact paths\u002Ffilenames requested, in the current workspace.\n- [ ] The processed media file opens, plays, and has a duration matching `compressed_duration_seconds` within a small tolerance.\n- [ ] JSON parses and contains every required key with the correct types (numbers as numbers, list of objects for segments).\n- [ ] Each removed segment has `start \u003C end`, `duration == end - start`, uses original-video time, and segments are sorted and non-overlapping.\n- [ ] `original_duration_seconds ≈ compressed_duration_seconds + removed_duration_seconds` within rounding tolerance.\n- [ ] `compression_percentage` matches `removed \u002F original * 100` (or the base specified) after rounding.\n- [ ] The opening static\u002Fnoise region is removed; spot-check that the first frames of output contain real content.\n- [ ] Long pauses above the stated threshold are removed while shorter, natural speech pauses are kept.\n- [ ] No values, filenames, thresholds, or code were copied verbatim from unrelated retrieved examples; all decisions trace to the target prompt.\n- [ ] Unrelated workspace files were not modified or deleted; temp files cleaned up.\n- [ ] Total processing time is comfortably under the stated runtime budget; a fallback fast-encode path was used if needed.\n","# SKILL.md\n\n## When to use\nUse this skill when the target task provides a media file (video\u002Faudio) and asks you to detect and remove low-information segments (silence, static openings, long pauses), then produce both a processed media artifact and a structured JSON report describing what was removed. The binding contract for this class of task typically requires:\n- A processed output file at a specified workspace path.\n- A JSON report with a fixed schema documenting original\u002Fcompressed\u002Fremoved durations, a compression percentage, and a list of removed segments with start\u002Fend\u002Fduration.\n- Internal arithmetic consistency (original ≈ compressed + removed).\n- Preservation of substantive content while trimming an opening and long pauses above some threshold.\n\nBefore acting, gather:\n- Exact input path, exact output filenames, and the exact JSON schema keys\u002Ftypes from the task prompt.\n- The pause-length threshold and any qualitative guidance about the opening (e.g., static frames, noise).\n- Time\u002Fcompute limits mentioned (e.g., must finish under a stated wall-clock budget).\n- Available tools in the environment (ffmpeg, ffprobe, Python audio libs).\n\n## Possible Failure Modes\n- Schema drift: renaming keys, nesting differently, or omitting required fields like `compression_percentage` or per-segment `duration`.\n- Inconsistent arithmetic: `original_duration_seconds` not ≈ `compressed_duration_seconds + removed_duration_seconds`, or `compression_percentage` computed against the wrong base.\n- Overly aggressive silence detection that cuts speech pauses shorter than the stated threshold, removing teaching content.\n- Overly conservative detection that leaves the static opening or long pauses intact, yielding too-low compression.\n- Detecting silence only by video (dark\u002Fstatic frames) or only by audio when the opening is characterized by both static video and noisy audio — missing one signal.\n- Reporting segments in the compressed timeline rather than in original-video time coordinates.\n- Merging adjacent silent segments incorrectly (or not at all), producing many tiny cuts and audible glitches.\n- Re-encoding with a slow codec \u002F high quality settings that blow the runtime budget, or using stream copy across cut points and getting keyframe-misaligned output.\n- Off-by-one\u002Fprecision issues: writing integer seconds when floats are expected, or float noise causing the sum check to fail.\n- Writing outputs to the wrong directory (not the current workspace) or with slightly different filenames than required.\n- Leaving temp files or partially written JSON if the pipeline errors midway.\n\n## Possible procedures\n1. Probe the input:\n   - Use `ffprobe` to get exact duration, streams, sample rate, and codec.\n   - Extract audio (mono, e.g., 16 kHz WAV) for analysis; keep the original for final cutting.\n\n2. Detect the opening:\n   - Use audio energy\u002FRMS over a sliding window and\u002For `ffmpeg -af silencedetect` from t=0 to find the first sustained speech onset.\n   - Optionally corroborate with a video signal (frame-difference or scene score) to confirm static frames at the start.\n   - Trim from 0 up to shortly before the first confident speech onset; avoid cutting into the first spoken words (leave a small pad, e.g., 0.2–0.5s).\n\n3. Detect long pauses:\n   - Run silence detection with a threshold and minimum-duration parameter matched to the task’s pause criterion (e.g., > the stated seconds).\n   - Sensible starting points: noise floor around -30 to -35 dB, min silence duration equal to (or slightly above) the task threshold.\n   - Merge overlapping\u002Fadjacent silence intervals; add a small keep-pad on each side so speech isn’t clipped.\n\n4. Build the removal plan in original-time coordinates:\n   - Produce a list of `[start, end]` segments to remove (opening + long pauses), sorted and non-overlapping.\n   - Compute keep segments as the complement within `[0, original_duration]`.\n\n5. Cut and concatenate:\n   - Prefer ffmpeg with a single filter graph (`select`\u002F`atrempo`-free `aselect` + `setpts`\u002F`asetpts` and `concat`) or generate segment files and concat-demux them.\n   - Re-encode with a fast but reasonable codec (e.g., libx264 `-preset veryfast`, AAC audio) to guarantee clean cuts and stay within the time budget.\n   - Verify the output file plays and has expected duration via `ffprobe`.\n\n6. Emit the JSON report:\n   - Use exactly the keys and nesting shown in the prompt.\n   - Populate durations from measured values: `original_duration_seconds` from ffprobe of input; `compressed_duration_seconds` from ffprobe of output; `removed_duration_seconds` = sum of removed segment durations; `compression_percentage` = removed \u002F original * 100 (confirm which base the prompt implies).\n   - Ensure each removed segment includes `start`, `end`, `duration` in original-video seconds, sorted ascending, non-overlapping.\n   - Round consistently (e.g., to 2–3 decimals) so the arithmetic check still holds.\n\n7. Sanity checks and recovery:\n   - If compression is implausibly high (e.g., > ~50%) or low (~0%), re-tune thresholds and re-run detection rather than shipping.\n   - If sum check fails, recompute from the segment list rather than patching numbers.\n   - If runtime approaches the budget, switch to faster ffmpeg presets or coarser analysis windows.\n\n## Verification Checklist\n- [ ] Both required output artifacts exist at the exact paths\u002Ffilenames requested, in the current workspace.\n- [ ] The processed media file opens, plays, and has a duration matching `compressed_duration_seconds` within a small tolerance.\n- [ ] JSON parses and contains every required key with the correct types (numbers as numbers, list of objects for segments).\n- [ ] Each removed segment has `start \u003C end`, `duration == end - start`, uses original-video time, and segments are sorted and non-overlapping.\n- [ ] `original_duration_seconds ≈ compressed_duration_seconds + removed_duration_seconds` within rounding tolerance.\n- [ ] `compression_percentage` matches `removed \u002F original * 100` (or the base specified) after rounding.\n- [ ] The opening static\u002Fnoise region is removed; spot-check that the first frames of output contain real content.\n- [ ] Long pauses above the stated threshold are removed while shorter, natural speech pauses are kept.\n- [ ] No values, filenames, thresholds, or code were copied verbatim from unrelated retrieved examples; all decisions trace to the target prompt.\n- [ ] Unrelated workspace files were not modified or deleted; temp files cleaned up.\n- [ ] Total processing time is comfortably under the stated runtime budget; a fallback fast-encode path was used if needed.\n",{"name":5,"description":7},"main","cdcb6d230f3b15c3c33f8a2bdd16439f43602cbe","evaluation\u002Fresults\u002Fskilltta-synthesis-opus47-tatumass-failed15-20260923\u002Ftasks\u002Fvideo-silence-remover\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\u002Fvideo-silence-remover\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",[],[]]