القائمة
Reconstruct a single event from multiple clips — sense each one, cross-anchor shared moments, surface corroborations and contradictions, and produce one cited chronological brief.
التثبيت باستخدام Codex أو Claude انسخ هذا Prompt والصقه في Codex أو Claude أو مساعد آخر ليراجع صفحة Skill ويثبّتها لك.
استنادا إلى تصنيف SOC المهني
Give any agent senses (video/audio/image understanding) and OSINT reach (search/capture/monitor) organized around an investigation case. Use when the user wants to analyze a video/audio/image, scan or monitor sources for a target, or ask/brief over accumulated findings. Drives the `overcast` CLI (built on pi + the tinycloud/Cloudglue perception backend); see reference/verbs.md for the full verb surface.
Install and configure overcast for this harness: install the CLI, verify the system ffmpeg, and configure reusable provider profiles. Use once per machine/profile before driving the `overcast` skill.
Turn a case into a visual evidence board — materialize face and object crops, link the same person across clips, connect themes, and render the corkboard as a CSI brief plus a live monitor wall.
Make unreadable footage legible — denoise/upscale a marked moment, re-run OCR and detection on the enhanced output, and record what was recovered with honest provenance (interpolation is a lead, not proof).
Work out where a photo or clip was taken — check embedded EXIF/GPS first, then pull signage, landmarks, and terrain clues, reverse-image-search the strongest ones, and corroborate to a location with cited evidence.
Run a standing surveillance watch on public sources for a target — auto-sense new media, auto-suggest matches into a triage queue, and keep a live control-room wall — so new evidence surfaces itself over time.
| name | overcast-timeline |
| description | Reconstruct a single event from multiple clips — sense each one, cross-anchor shared moments, surface corroborations and contradictions, and produce one cited chronological brief. |
Use this skill to "walk through what happened" across several recordings of one
event (bystander videos, multiple cameras, a sequence of clips): build one ordered,
cited timeline and flag where accounts disagree. Use the broad overcast skill and
overcast/reference/verbs.md for exact flags.
watch for the visual timeline, listen where audio carries
the account:overcast doctor --json
overcast case init --json
overcast watch ./cam1.mp4 --json
overcast watch ./phone-clip.mp4 --json
overcast listen ./phone-clip.mp4 --json
overcast note "wall clock reads 21:14 as the door opens" --ref <cam1-record-id> --at 12-15 --tag anchor --json
overcast note "same doorbell chime as cam1@13 — clips overlap here" --ref <phone-record-id> --at 3-6 --tag anchor --json
overcast ask "order the events across all clips with timestamps; where do accounts agree or conflict? cite record.id + media.at" --json
# only if the answer reports a real conflict:
overcast finding create "conflict: cam1 shows the car arriving BEFORE the shout; phone-clip audio has the shout first" --ref <cam1-record-id> --at 12-15 --confidence low --json
tldr note first:overcast note "reconstructed <n> clips into one timeline; <k> firm anchors; <c> unresolved conflicts" --tag tldr --json
# Wait for the note result before exporting, so the TL;DR is included.
overcast brief --export ./timeline.html --json
One ordered chronology of the event, each entry cited to a record.id + media.at,
the anchor moments that let clips be aligned, and an explicit list of unresolved
contradictions. Where clips can't be ordered relative to each other, say so rather
than guessing.
Device clocks and upload times drift — prefer content anchors (a shared sound, a
visible clock, a synchronized event) over file timestamps when aligning clips.
Absence of a moment in one clip is not evidence it didn't happen — it may be
off-frame. Keep observed facts (in note) separate from inferred ordering; a
contradiction is a finding to review, not a settled conclusion.