Add focused NeMo Retriever agent skills

[jeremyjordan]

Description

This PR adds a focused NeMo Retriever agent-skills bundle produced through agent-driven exploration of the repository and validation against the current CLI surface. The skills are scoped to supported or practically usable Retriever workflows, packaged under nemo_retriever/.agents/skills, exposed through per-skill symlinks in .agents/skills and .claude/skills (so they can coexist with other contributor-focused root skills), and included in the wheel package data.

The PR also keeps the CLI guidance honest for agents by labeling top-level subcommands that are experimental and not product-supported in retriever --help and the corresponding subcommand help pages.

Skills Added

• nemo-retriever-setup: installation, environment verification, extras, model/API prerequisites, and first CLI checks.
• nemo-retriever-ingest: one-shot ingest, pipeline ingest, LanceDB table creation, embedding configuration, and ingest validation.
• nemo-retriever-query: querying existing LanceDB indexes, matching ingest settings, reranking behavior, hit-schema inspection, and evidence-grounded answer synthesis.
• nemo-retriever-service: local service startup, service ingest client usage, HTTP endpoint troubleshooting, config handling, and Helm/service deployment orientation.

Other Changes

• Replaces the previous single root .agents symlink with per-skill symlinks in .agents/skills/ and .claude/skills/.
• Removes the older monolithic .claude/skills/nemo-retriever skill.
• Adds CLI/test fixes discovered while exercising the skills against real commands, including local HuggingFace query flags and split-skill prompt handling. I'm including the changes in this PR because the skills depend on them, but they should likely be separated out into a different focused MR before merging.
• Adds root CLI help labels for experimental, not product-supported command groups.

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.
• If adjusting docker-compose.yaml environment variables have you ensured those are mimicked in the Helm values.yaml file.

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.