Add Nemotron Custom Policy Generator skill#237
Open
svenchilton wants to merge 1 commit into
Open
Conversation
Author
|
CI failure is due to 2 pre-existing broken links in train-models/reference/optimize/prune.md and distill.md — not related to this PR. The links (docs.nvidia.com/nemo/megatron-bridge/latest/training/pruning.html and distillation.html) return 404 on main as well. @rkalaniNV, @chrisalexiuk-nvidia, or whoever else is appropriate, would you be kind enough to approve this PR anyway, or tell me how to fix the broken links? |
rkalaniNV
requested changes
May 30, 2026
Collaborator
rkalaniNV
left a comment
rkalaniNV
left a comment
There was a problem hiding this comment.
Thanks for contributing this skill.
- Please remove Computex/future-launch references from SKILL.md and prompt templates
- Please make the skill name/folder/path consistent everywhere
- Please remove internal publishing/audit/process notes from user-facing files.
- Please align taxonomy mode names between SKILL.md and policy_json_schema.json
Once done, I can trigger the nvskills-ci bot to give you more insights as well as populate BENCHMARK.md
| Run it twice, once for each agent harness the catalog supports (Claude Code and Codex), per the publishing guide: | ||
|
|
||
| ```bash | ||
| nv-aces run --skill skills/nemotron-policy-generator \ |
Collaborator
There was a problem hiding this comment.
The folder name is nemotron-custom-policy-generator. Can you please check if this is intentional here. This can affect discoverability of the skill.
| @@ -0,0 +1,126 @@ | |||
| # Publishing Audit — nemotron-policy-generator | |||
| ownership: # team-owned | ||
| team: "Nemotron Safety PM" | ||
| primary_contact: "Shyamala Prayaga <sprayaga@nvidia.com>" | ||
| slack_channel: "#nv-skills-onboarding" |
Collaborator
There was a problem hiding this comment.
Can we remove the internal slack channel for public repo?
| @@ -0,0 +1,138 @@ | |||
| ## Description: <br> | |||
Collaborator
There was a problem hiding this comment.
Why do we need multiple skill card files?
| --- | ||
| name: "nemotron-policy-generator" | ||
| title: "Nemotron Custom Policy Generator" | ||
| description: "Generates BYO custom safety policies for NVIDIA Nemotron content-safety guardrails. Targets two Gemma-3-4B fine-tunes: nvidia/Nemotron-Content-Safety-Reasoning-4B (text, English, /think and /no_think, BYO today) and nvidia/Nemotron-3-Content-Safety (multimodal text+image, 12 languages, BYO + reasoning at Computex 2026). Both share the Nemotron Content Safety V2 taxonomy. Accepts loose keywords, keywords + deployment context, an existing policy to extend, or free-form prose. Produces a Markdown policy, JSON taxonomy, drop-in inference prompts for the chosen model+mode, and optional .docx. Maps rough words to V2 canonical categories; adds custom categories or topic-following rules where needed. Use whenever the user mentions building, drafting, generating, or extending a safety policy, content policy, moderation policy, guardrail config, BYO-policy, custom safety taxonomy, eval rubric, labeling rubric, or topic-following rules — even when they just say 'rough words for a Nemotron policy'." |
Collaborator
There was a problem hiding this comment.
Computex references in skill file.
| "type": "array", | ||
| "items": { "enum": ["runtime_guardrails", "training_data_labeling", "customer_byo_policy", "eval_rubric"] } | ||
| }, | ||
| "taxonomy_mode": { "enum": ["clean_aegis", "aegis_plus_custom", "mostly_custom"] }, |
Collaborator
There was a problem hiding this comment.
Diff from skill.md where clean_v2 is mentioned.
Generates deployable content safety policies for NVIDIA Nemotron content safety models from natural-language safety requirements. Supports dual-target: Nemotron-Content-Safety-Reasoning-4B (text, English) and Nemotron-3-Content-Safety (multimodal, 12 languages). Skill authored by Shyamala Prayaga; repo restructure by Sven Chilton. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com> Signed-off-by: Sven Chilton <schilton@nvidia.com>
svenchilton
force-pushed
the
schilton-publish-nemotron-custom-policy-generator-skill
branch
from
0e9e33b to
afc49a1
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Generates deployable content safety policies for NVIDIA Nemotron content safety models from natural-language safety requirements. Supports dual-target: Nemotron-Content-Safety-Reasoning-4B (text, English) and Nemotron-3-Content-Safety (multimodal, 12 languages).
Skill authored by Shyamala Prayaga; repo restructure by Sven Chilton.