Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add calesthio/generative-media-skills --skill video-description-oversightgit clone --depth 1 https://github.com/calesthio/generative-media-skillsWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/calesthio/generative-media-skills/video-description-oversight)<a href="https://agentmods.dev/skills/calesthio/generative-media-skills/video-description-oversight"><img src="https://agentmods.dev/badge/skills/calesthio/generative-media-skills/video-description-oversight/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/calesthio/generative-media-skills/video-description-oversight"><img src="https://agentmods.dev/badge/skills/calesthio/generative-media-skills/video-description-oversight.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00086 | $0.02744 |
| Opus 5 | $0.00043 | $0.01372 |
| Sonnet 5 | $0.00017 | $0.00549 |
| Haiku 4.5 | $0.00009 | $0.00274 |
Grade A, and why
video-description-oversight scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 12d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 269 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Video description oversight
Use this skill when language about a video must be accepted as evidence-quality metadata rather than plausible prose. It governs a correction loop:
source video + specification + pre-caption
-> evidence-backed critique
-> revised post-caption
-> independent acceptance review
-> versioned triplet and decision record
The reviewer verifies what language says about media. Technical file integrity, visual artifact QA, accessibility captions, creative intent, and model post-training belong elsewhere.
Evidence stance
- Documented fact: a finding supported by the source video, approved specification, or cited standard/research.
- Reviewer observation: a timestamped comparison between video and description.
- Production heuristic: an operational quality-control choice that must fit the project's risk and workforce.
The workflow is informed by Lin et al.'s CHAI framework, which reports that critiques used to revise precise video captions are most useful when accurate, complete, and constructive. Sources were verified 2026-07-14. CHAI's performance numbers are first-party research results on its data and annotator program, not universal service-level guarantees.
Activation and boundaries
Use this skill for descriptions used in:
- search, archive, edit logging, or reference analysis;
- training/evaluation datasets;
- generation-prompt handoff;
- cinematography-aware video understanding;
- regulated or high-risk metadata where language errors matter.
Route elsewhere for:
- codec, color, audio, anatomy, artifact, or final-media release QA;
- SDH/subtitle timing and accessibility compliance;
- deciding the desired shot or creative treatment;
- training/fine-tuning reward or caption models;
- provider-specific video analysis API calls;
- payroll, labor classification, or workforce procurement.
The workflow requires a description specification. If none exists, establish one through precise-video-description before grading completeness or terminology.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 12d ago First seen · 269 lines · 86 tokens per session scan A 87bed5ade470
video-description-oversight is a skill published in the GitHub repository calesthio/generative-media-skills (170 stars, last pushed 1mo ago), licensed MIT. It adds 86 tokens to every session and 2,744 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
cliptalk-audio-polish-mixer
Produces a task-local audio-polished preview or output from an existing cut, applying loudness normalization, optional noise reduction, fades, muting, and voice-first mix policy.
cliptalk-broll-overlay-editor
Adds relevant B-roll or cutaway overlays to an existing timeline while preserving the primary audio and making the result reviewable before final export.
cliptalk-cover-intro-composer
Creates task-local cover candidates and composes the confirmed cover into the beginning of the current output as a short intro, keeping cover selection separate from video editing.
feature-demo-recording
Record a demo video of a web feature from a real browser. Two modes -- a NARRATED film where measured voiceover drives the timeline (designed slides, subtitles, punch-in camera, rendered from an HTML timeline), and a SILENT evidence clip for a PR or a QA pass. Use when the user asks to record a video, demo, or screen…
image-authoring
Author images and diagrams as code — SVG, Pillow, Excalidraw, mermaid. Load when asked to draw, illustrate, or make an image, icon, logo, poster, or diagram.
bento-slides
Create and edit Bento presentations — self-contained .bento.html decks whose document is JSON. Use whenever the user wants a slide deck or presentation: from scratch, from source material, or by improving an existing file.