| name | qa-tech-specs |
| description | Validates a generated piece against platform technical specs for aspect ratio, resolution, duration, file size, and codec |
| version | 0.1.0 |
QA Tech Specs
Pre-publish validator for asset technical compliance. Reads platform specs from .specs/product/CHANNELS.md and compares them to the actual file metadata. Blocks publishing when a critical spec is violated.
When to invoke
- Right before scheduling any asset on
adaptlypost or any direct platform API.
- After image or video generation by Higgsfield, Topview, Wavespeed, or gpt-image.
- When re-purposing an asset across platforms (a 1:1 IG image being pushed to a 9:16 Reel).
- During QA on a sprint piece in
.specs/sprints/sprint-XX/.
- When a publish attempt fails with a platform-side spec error and we need to confirm the asset shape.
Inputs
asset_path: string. Local path or URL to the file.asset_kind: string. image, video, carousel, audio.target_platforms: array of strings. Subset of ig_feed, ig_reel, ig_story, tiktok, youtube_shorts, youtube_long, facebook_feed, linkedin, x.channels_md_path: string, optional. Override path to .specs/product/CHANNELS.md.
Process
- Read
channels_md_path (default .specs/product/CHANNELS.md) and parse the spec table for each platform: aspect ratio, min and max resolution, max duration, max file size, accepted codec, accepted container.
- Probe the asset file metadata using
ffprobe for video and identify for image. Extract aspect ratio, resolution, duration, file size, codec, container.
- For each
target_platform, compare actual values against the spec and record violations.
- Classify violations:
hard (platform will reject) versus soft (platform will accept but quality drops).
- Build a per-platform pass map with violation details and suggested fixes (re-encode, crop, scale).
- If any
hard violation appears, set the overall pass flag to false.
- Return the report.
Outputs
pass: boolean. False if any hard violation exists.per_platform: object. Map of platform -> { pass, violations, fixes }.metadata: object. The actual file metadata extracted.
Examples
Example 1: 1080x1080 image targeting IG feed and IG Reel
Input: { asset_path: "outputs/quote-card.png", asset_kind: "image", target_platforms: ["ig_feed", "ig_reel"] }
Output: { pass: false, per_platform: { ig_feed: { pass: true, violations: [], fixes: [] }, ig_reel: { pass: false, violations: [{ rule: "aspect_mismatch", expected: "9:16", actual: "1:1", severity: "hard" }], fixes: ["regenerate at 1080x1920 or center-crop with safe area"] } }, metadata: { width: 1080, height: 1080, format: "png", file_size_kb: 412 } }
Example 2: 30s vertical video targeting TikTok and YouTube Shorts
Input: { asset_path: "outputs/reel-001.mp4", asset_kind: "video", target_platforms: ["tiktok", "youtube_shorts"] }
Output: pass true, both platforms accept the file.
Failure modes
- ffprobe or identify not installed: surface the missing tool and the install command.
- CHANNELS.md missing the target platform: log a warning, skip that platform, and recommend updating the spec doc.
- Asset path unreachable: throw a descriptive error with the resolved path.
- File codec accepted but at the edge of the spec: mark as
soft and recommend a re-encode.
Related skills
caption-multi-platform: caption side of the same publish gate.wavespeed-batch: produces batch assets that should be validated by this skill.topview-prompt-builder: produces UGC video that often needs spec validation.higgsfield-prompt-builder: produces cinematic video that may need a re-encode pass.
