asking-with-evidence
Query any video you've already watched to get answers backed by timestamps and confidence scores, drawn from a persistent index rather than re-processing. The skill refuses unclear answers honestly and escalates automatically when uncertain, saving tokens in the process.
Asking with Evidence answers questions about previously watched videos by searching a persistent index with timestamped results.
AI-generated summary based on this skill's SKILL.md
Install
oxbshw/watch-skill/asking-with-evidence · repository language: Python
git clone https://github.com/oxbshw/watch-skill
cp -r watch-skill/adapters/claude-skill/skills/asking-with-evidence ~/.claude/skills/asking-with-evidencenpx skillfed install oxbshw/watch-skill/asking-with-evidenceFrequently asked questions
AI-generated answers based on this skill's SKILL.md and metadata
What did they say about X in the video?
asking-with-evidence lets you query what was said about any topic in a video you've already watched. The skill searches its persistent index and returns the exact transcript excerpt along with the timestamp where it appears and a confidence score, so you have proof of what was stated without rewatching.
How does asking-with-evidence find error codes with timestamps?
asking-with-evidence indexes all watched videos and can locate specific error codes or technical details at their exact moments. When you ask for an error code, it returns the timestamp, confidence score, and surrounding context so you can jump directly to that part of the video.
Can I search all my watched videos at once?
Yes. asking-with-evidence supports semantic and keyword search across your entire library of watched videos. Instead of re-watching or manually scanning, you query the indexed content and get back matching moments with timestamps and confidence scores for each result.
What happens when asking-with-evidence is uncertain?
asking-with-evidence refuses unclear answers honestly and escalates automatically when confidence is too low. This prevents hallucination, saves tokens by avoiding re-processing, and ensures you only get answers backed by real evidence from the video index.
How do I get transcript excerpts with exact timestamps?
asking-with-evidence extracts transcript excerpts and frames at exact timestamps as evidence. Query the video by topic, phrase, or moment, and the skill returns the relevant text or frame along with its precise timestamp and confidence score.
Can asking-with-evidence answer follow-up questions without rewatching?
Yes. asking-with-evidence maintains a persistent index of watched videos, so follow-up questions are answered by querying that index rather than re-processing the video. This saves time and tokens while keeping your answers consistent and traceable to specific timestamps.
SKILL.md
rendered from the published skill — quoted content, verbatim
Asking with evidence
Every watched video sits in a persistent index. Questions about it are answered from that index — text first, frames only when needed — with timestamps, a confidence score, and an honest refusal when the video does not show the answer. Never re-run a watch for a follow-up.
Answer a question
watch-skill ask <video_id-or-original-url> "<question>"
Any language works; the answer comes back in the language of the
question. The engine escalates on its own when unsure (dense re-sampling,
zoom-crop re-OCR, stronger model) and prints a ~N tokens saved line.
Three rules for reading the result:
- Cite the timestamps it gives you; they are real evidence, not decoration.
- Trust the refusal. When it says the
(truncated - see the full file via the links below)
Read as markdown · JSON record · Browse the source repository
File tree — 1 file
adapters/claude-skill/skills/asking-with-evidence/SKILL.md