AGENT SKILL

analyze-footage

Analyze one footage file in pull-driven levels L0–L3 and record only observed evidence in analysis.json v0. By default, run L0 metadata and L1 audio (`akari media probe` → `waveform` → `transcribe` only when speech is likely); cloud transcription requires approval. Inspect keyframes, editing events, and hooks at L2, or person tracks at L3, only when requested. Use for new footage, pre-edit analysis of one file, or footage analysis needed before plan-edit.

License 许可
MIT
最近更新
2026-10-22
Stars
★ 214
Forks
25

01 前言 Front Matter

name
analyze-footage
description
Analyze one footage file in pull-driven levels L0–L3 and record only observed evidence in analysis.json v0. By default, run L0 metadata and L1 audio (`akari media probe` → `waveform` → `transcribe` only when speech is likely); cloud transcription requires approval. Inspect keyframes, editing events, and hooks at L2, or person tracks at L3, only when requested. Use for new footage, pre-edit analysis of one file, or footage analysis needed before plan-edit.

02 资源文件 63

03 SKILL.md 正文

Analyze one footage file

Language: Respond in the user's language — conversation, questions, approval prompts, and the reports and records you write (including decision-log rows) follow the language the user writes in.

Read lang/<content-language>.md for language-specific rules (content language: --lang → intake.json "language" → transcript language → ja; see https://github.com/AkariLabs/akari-video/blob/main/docs/contract-2026-10-11-content-language.md). If that file does not exist, read lang/en.md.

Analysis levels

The authoritative source for terminology, inputs, outputs, and ledger updates is akari media Observation Command Contract v0 §3 / §4. Use akari media to observe the source. This skill interprets and scores evidence, then records only observed results.

Level One-line summary
L0 metadata Use probe to confirm duration and available video and audio streams.
L1 audio Run waveform first; run transcribe only when speech_likely is true.
L2 visuals Visually inspect filmstrip / grab images and record only needed keyframes, events, and hooks.
L3 people Generate only the person sidecars needed for the chosen effect and time range.

Default to L0 + L1. Add L2 or L3 only when the request explicitly calls for it or a later skill asks for a specific observation. A requested level is a ceiling, not an obligation to fill every section.

Hard rules

  • Analyze one footage file per run, recording every time in seconds from the uncut source.
  • Never modify the source. Keep outputs at the canonical destination in workflow.md to prevent collisions between files with the same name.
  • Invoke akari media <sub> <target> [options] under the akari media contract §1 / §2. Parse stdout as JSON or JSON Lines; do not call a media backend directly from this workflow.
  • Create a 720p proxy only if visual review at L2 or above needs it. At L0 and L1, pass the original file to probe, waveform, and transcribe without making a proxy.
  • Follow contract §2.5 for transcription backend selection. A cloud connection must be registered in .akari/connections.json with doctor status ok, explicitly selected, and approved by a person through a decision card. Never upload audio by default. Access keys only through credentials.env and never print their values in chat, artifacts, or logs.
  • Obtain word timing through akari media transcribe, which snaps word starts to speech after silence. Do not use word times from a direct whisper-cli call: its default offsets are estimates and place words before speech when initial or inter-sentence silence is present (#92 observed a first word at zero despite two seconds of silence). With flash attention enabled by default, -dtw <preset> is silently ineffective unless used with -nfa and read from tokens[].t_dtw.
  • If no backend is available or no speech is present, do not invent words. Leave transcript: [] and report why.
  • Write note after actually viewing the images used in L2. Do not write findings on unseen images.
  • Create only filler | trouble | chapter | highlight | hook events. Record a hook only for a window chosen for scoring, and score all five axes there with integers from 1 to 5.
  • Add no fields outside analysis.schema.json. Finalize only JSON that passes structural and semantic checks. keyframes: [] and events: [] are valid both when L2 was not observed and when nothing qualified; distinguish the cases by the type and range of observations[].
  • Do not generate L3 for every file. Generate only the person-track types and intervals needed for the chosen effect. By default, tracks.person_matte: null; omit optional person-track keys until generated.
  • Use OpenMontage for structural reference only; do not copy AGPL text or code.
  • Follow the "existing contract sidecars" layer of the project-structure-v0 contract for .akari/sidecars/ output paths.

Pass --lang for the work's content language on transcription calls. The default SpeechAnalyzer backend uses that language when its locale is supported.

Execution order and stopping conditions

  1. Read workflow.md to confirm the source, output destination, and requested level.
  2. L0 — probe: Run akari media probe. You may stop if the request asks only for L0, or there is no audio for L1 to inspect.
  3. L1 — waveform: If audio exists, run akari media waveform. You may stop if speech_likely is false, or speech content and subtitles are not needed and transcription was not explicitly requested.
  4. L1 — transcribe: Run akari media transcribe only when speech_likely is true. You may stop once the default L0 + L1 work is complete and no L2 or L3 work was requested. If no backend is available, stop with an explanation of the degraded result rather than guessing.
  5. L2 — visuals and events: Only when requested, read keyframes-and-review.md and events-and-hooks.md, then inspect and judge the necessary windows. You may stop when the requested keyframes, events, or hooks are covered and no person track is needed.
  6. L3 — people: Only when requested, run the relevant type and interval from person-matte.md or vision-tracks.md. You may stop when the selected effect has its inputs, or when a sidecar is unavailable and the reason is recorded.
  7. Read analysis-json.md. Preserve untargeted existing sections, validate against the schema and semantic rules, then finalize atomically. Validation is required at every stopping level.

Read only the files needed for the current step and requested level.

04 同仓库更多 Skills

AkariLabs
AkariLabs/akari-video

verify-project

Verify an AKARI Video project with the L0 / L1 / L2 verification ladder required by an AKARI Video task contract on the current Theia stack. Read when acceptance criteria specify levels such as L0 or L0+L1, to determine exactly what to run and how at each level.(旧名: verify)

★ 214
AkariLabs
AkariLabs/akari-video

verify-edit

Verify edit.json and optional analysis.json / captions.json / media with a deterministic CLI, then finish QA with frame inspection and reporting after PASS. Use immediately after writing or modifying edit.json, before export, and after applying review feedback.(旧名: edit-lint)

★ 214
AkariLabs
AkariLabs/akari-video

setup-remote

Set up remote access to approval reports and previews from another device such as a phone, and send footage to the workspace. Use doctor to assess Tailscale, guide installation and login (human actions), expose the preview server (default 4567) and approval report helper over tailnet-only HTTPS with tailscale serve, connect Taildrop reception to workspace inbox/, and verify access from another device. Use for requests to view reports on a phone, approve remotely, send footage from a phone, or set up remote access. Public internet exposure (funnel) is outside the default scope.

★ 214
AkariLabs
AkariLabs/akari-video

setup-project

Set up an AKARI Video project headlessly. Recursively copy `templates/project-default/`, record the template version, initialize git only when safe, and generate an HTML creation report. No app launch is needed. Use to create a video project or fill in an existing folder as an AKARI Video project.(旧名: create-project)

★ 214
AkariLabs
AkariLabs/akari-video

setup-library

Use for first-time AKARI Video setup or when the current project needs more usable assets. This first-run skill checks ffmpeg / whisper-cli (and can install with approval), reads the bundled catalog with akari assets to propose a starter pack, obtains explicit human approval, then acquires, places, validates, and updates INDEX entries. It also covers the audio library (BGM and sound effects): bulk acquisition of first-party AKARI Sounds from GitHub Releases, a free-source candidate HTML list with drop-folder matching or assisted acquisition for missing families, and keep/drop decisions in the listening gallery. Use for requests such as "I want more BGM" or "add sound effects to the library".(旧名: setup-audio-library)

★ 214
AkariLabs
AkariLabs/akari-video

setup-connections

Set up and inspect AKARI Video connections: AI-service keys, which models are allowed, cost-approval policy, the tool and media-service state, and 3D settings. Use for first-time setup, adding a provider or key later, checking what is usable (「何が使える?」 'what can I use?'), setting a key such as fal (「fal の鍵を設定したい」), 3D settings (「3D の設定を使いたい」), connection checks, and gates before paid generation or external publication. Model selection belongs to the production skills (they read the ledger and preferences); this skill decides allowed models and the cost policy. Works with `.akari/connections.json` and a free, read-only doctor.(旧名: manage-connections)(旧名: connect)

★ 214