agentleFS
45ck45ck/video-evaluator/skills/video-artifact-intake/SKILL.md

video-artifact-intake

Skill2 starsChanged 7 months agoMITView on GitHub

Normalize a local video run into a consistent artifact bundle by resolving output roots, latest pointers, known reports, and the primary video path.

The file

We split this skill into the rules it gives an agent

5 rules. About 0 are common, 2 are distinctive, and 3 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 3

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this when an agent needs to understand what a repo produced before reviewing or comparing it.

    line 8
  2. Very specific:

    Use outputDir when you already know the run directory.

    line 31
  3. Very specific:

    Run this before review-bundle, package-review-prompt, or compare-video-runs if you are unsure which directory or video should be reviewed.

    line 58
  4. Very specific:

    When timeline.evidence.json is present, inspect it before making sequence, caption, transcript, or action-timing claims.

    line 63
  5. Distinctive:

    Do not infer product behavior, UI quality, or correctness from intake alone.

    line 78

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Video Artifact Intake

Use this when an agent needs to understand what a repo produced before reviewing or comparing it.

This skill owns:

  • resolving latest.json or LATEST.txt
  • finding quality.json, verification.json, validate.json, score.json, publish.json, and other common artifacts
  • locating the main MP4/WebM artifact
  • probing the video with ffprobe when available

Inputs

Provide one or more bundle locators:

{
  "outputDir": "./output/example-run",
  "latestPointerRoot": "./output",
  "videoPath": "./output/example-run/output.mp4"
}

Use outputDir when you already know the run directory. Use latestPointerRoot when the repo maintains latest.json or LATEST.txt. Use videoPath when the video is outside the normal output directory. Paths may be relative to the current working directory.

Outputs

The result is a normalized bundle map with:

  • rootDir: resolved run directory, or null if only a standalone video was provided.
  • videoPath: primary video path if found.
  • artifacts: known reports and files found under the root.
  • reportStatuses: derived status for JSON reports with status, passed, or hasFailures fields.
  • overallStatus: pass, warn, fail, or unknown.
  • recommendedFocus: artifact-driven review areas.
  • videoProbe: duration, dimensions, codec, and size when ffprobe succeeds.
  • timeline.evidence.json: generated when timestamps.json, events.json, or subtitles.vtt exists in the bundle.
  • video.shots.json: discovered when shot extraction has already been run for the bundle.
  • segment.evidence.json: discovered when segment fusion has already been run for the bundle.

Sequencing Guidance

Run this before review-bundle, package-review-prompt, or compare-video-runs if you are unsure which directory or video should be reviewed. If storyboard artifacts are missing but a video is present, continue with storyboard-extract.

When timeline.evidence.json is present, inspect it before making sequence, caption, transcript, or action-timing claims. When video.shots.json is present, use it as a coarse part map before deep visual review. When segment.evidence.json is present, inspect it first for per-segment evidence availability and abstention signals.

Interpretation Notes

overallStatus: unknown usually means no recognizable JSON status reports were found; it is not a pass. reportStatuses are shallow summaries and do not replace reading the source reports when they warn or fail. videoProbe is optional because ffprobe may be unavailable or unable to parse the file.

Abstention Rules

Do not infer product behavior, UI quality, or correctness from intake alone. Abstain from quality judgments if the only evidence is a discovered video path and no review reports or storyboard artifacts.

Failure Modes

Missing or stale latest pointers can resolve the wrong run. If no video is found, check whether the repo uses nonstandard filenames and pass videoPath explicitly. If ffprobe is missing, the bundle can still be mapped but video metadata will be absent.

Repo-side runner:

node --import tsx scripts/harness/video-intake.ts

Installed-pack runner:

node ./.video-evaluator/agent/run-tool.mjs video-artifact-intake

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.