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 NVIDIA-TAO/tao-skill-bank --skill paidf-cosmos-predictgit clone --depth 1 https://github.com/NVIDIA-TAO/tao-skill-bankWrote 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/nvidia-tao/tao-skill-bank/paidf-cosmos-predict)<a href="https://agentmods.dev/skills/nvidia-tao/tao-skill-bank/paidf-cosmos-predict"><img src="https://agentmods.dev/badge/skills/nvidia-tao/tao-skill-bank/paidf-cosmos-predict/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/nvidia-tao/tao-skill-bank/paidf-cosmos-predict"><img src="https://agentmods.dev/badge/skills/nvidia-tao/tao-skill-bank/paidf-cosmos-predict.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 6 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Privilege Escalation · line 151 Potential security issue detected. Manual review is recommended.Fix: Review the flagged content for security risks. Ensure no credentials, secrets, or sensitive data are exposed.
- high Privilege Escalation · line 152 Potential security issue detected. Manual review is recommended.Fix: Review the flagged content for security risks. Ensure no credentials, secrets, or sensitive data are exposed.
- medium MCP Rug Pull · line 149 Docker image references without a specific tag (:latest is implicit) or digest (@sha256:...) can be silently replaced by a malicious image.Fix: Pin the image: image:tag or image@sha256:abc123
- medium Excessive Agency · line 239 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- medium Excessive Agency · line 240 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
- medium MCP Rug Pull · line 243 Docker image references without a specific tag (:latest is implicit) or digest (@sha256:...) can be silently replaced by a malicious image.Fix: Pin the image: image:tag or image@sha256:abc123
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 | $0.00023 | $0.04366 |
| Opus 5 | $0.00012 | $0.02183 |
| Sonnet 5 | $0.00005 | $0.00873 |
| Haiku 4.5 | $0.00002 | $0.00437 |
Grade A, and why
paidf-cosmos-predict 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 — 248 lines — stays where its author put it; the contents beside it link to each section on GitHub.
PAIDF Cosmos Predict Generation
Prepare and run PAIDF Cosmos Predict generation for media samples. The skill emits a JSONL handoff that maps each input id to the original media path and generated video path.
Purpose
Use this skill when a DEFT workflow already has media samples and needs synthetic/generated videos from PAIDF Cosmos Predict. This skill does not start the VLM captioning service. A reachable OpenAI-compatible base URL for the model used to caption input media must be provided at runtime.
Prerequisites
- Docker with NVIDIA GPU support and
nvidia-container-toolkit. - Access to the PAIDF augmentation image declared by
images.metropolis_sdg.paidf_augmentationinversions.yaml. - A running VLM captioning service with an OpenAI-compatible API base URL. The base URL must be provided by the user or upstream workflow at runtime; reuse that exact base URL for every
--vlm-captioning-endpointargument. Do not include/modelsinVLM_CAPTIONING_ENDPOINT. HF_TOKENin the run environment when Cosmos model downloads require HuggingFace access — exported, or in a user-approved env file (bareKEY=valuelines) that the run block sources.VLM_API_KEYin the run environment the same way when the VLM captioning endpoint requires authentication.- Input media paths that are absolute paths on the host under the required media directory. Pass the host media directory with
--media-dir; the skill mounts it into the PAIDF container at the exact same path.
Inputs
| Input | Required | Notes |
|---|---|---|
| Input JSONL | Yes | Path to the generic media JSONL. The user or upstream workflow must provide it. Each row must include string fields id and media_path; id values must be unique. |
| Output directory | Yes | Host directory for prepared PAIDF config, generated videos, captions, metadata, logs, and final handoff. The user or upstream workflow must provide it. |
| VLM captioning endpoint base URL | Yes | The user or upstream workflow must provide this OpenAI-compatible base URL, for example a URL ending in /v1. Do not include /models. Pass the exact same base URL to verify_vlm_captioning_base_url.py before any other step and to prepare_paidf_config.py when writing config.yaml. |
| Generation settings | No | If the user provides a generation settings JSON, use it. Otherwise set GENERATION_SETTINGS to skills/data/paidf-cosmos-predict/assets/default_generation_settings.json. Always pass the resolved path to prepare_paidf_config.py with --generation-settings. |
| PAIDF GPU count | Yes | Number of GPUs for PAIDF augmentation. Pass it to prepare_paidf_config.py with --paidf-num-gpus N and to Docker with --gpus "$PAIDF_NUM_GPUS". |
| Media directory | Yes | Host directory containing the input media referenced by media_path. Pass it as --media-dir /path/to/media_dir to prepare_paidf_config.py and mount it into Docker 1:1. |
| VLM captioning prompt | Yes | Prompt text file for VLM captioning. Pass it to prepare_paidf_config.py with --caption-prompt-file; the script inlines the prompt into config.yaml. |
HF_TOKEN |
Yes for Cosmos model downloads | The agent checks that HF_TOKEN is already set in the run environment, whether exported or sourced from a user-approved env file, and forwards it to Docker with -e HF_TOKEN. |
VLM_API_KEY |
Yes when the VLM captioning endpoint requires authentication | The agent warns when VLM_API_KEY is not set, then forwards it to Docker with -e VLM_API_KEY when present. If the endpoint does not require authentication, PAIDF can run without it. |
What ships with it
8 files 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.
- assets/default_generation_settings.json 728 B
- assets/paidf_config_template.yaml 1.7 KB
- evals/evals.json 807 B
- scripts/paidf_common.py 12 KB runs code
- scripts/prepare_paidf_config.py 5.1 KB runs code
- scripts/test_paidf_cosmos_predict.py 17 KB runs code
- scripts/verify_vlm_captioning_base_url.py 962 B runs code
- scripts/write_paidf_handoff.py 2.9 KB runs code
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 · 248 lines · 23 tokens per session scan A 85d5cded4775
paidf-cosmos-predict is a skill published in the GitHub repository NVIDIA-TAO/tao-skill-bank (88 stars, last pushed today), licensed Apache-2.0. It adds 23 tokens to every session and 4,366 once invoked, about $0.0001 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
seedance-2.0-prompter
Expert prompt engineering for Seedance 2.0. Use when the user wants to generate a video with multimodal assets (images, videos, audio) and needs the best possible prompt.
multimodal-llm
Vision, audio, video generation, and multimodal LLM integration patterns. Use when processing images, transcribing audio, generating speech, generating AI video (Kling v3, Sora 2, Veo 3.1 std/lite/fast, Runway Gen-4.5 via gen4turbo), or building multimodal AI pipelines.
replicate-runner
Use when run AI models on Replicate cloud API. Deploy image generation, video creation, audio processing, and custom models without managing infrastructure. Use when working with replicate runner.
ltx-video
Set up, configure, and run LTX-2/LTX-2.3 (Lightricks) for AI video and audio generation. Use when: (1) Installing LTX-2 locally or via ComfyUI, (2) Generating video from text, image, audio, or keyframes, (3) Running inference with the 22B dev/distilled models, (4) Fine-tuning with LoRA/IC-LoRA, (5) Configuring…
data-designer
Use when the user wants to create a dataset, generate synthetic data, or build a data generation pipeline.
kilocode-multimodal
Verify tool execution using text, audio, and VLM checks across PMOVES services on the KiloCode GLM lane. Use when validating multimodal outputs, capturing verification evidence, or binding verification to task/run IDs.