CtrlK
BlogDocsLog inGet started
Tessl Logo

nemo-rl-auto-research

Autonomous NeMo-RL research agent workflow for directed hypothesis testing and open-ended discovery. Guides agents through the full experiment lifecycle: understanding recipes and environments, wiring RL or NeMo-gym runs, launching reproducible baselines and iterations, analyzing results, preserving human oversight, and using git plus TSV logs as the research ledger. Do NOT use for: bug fixes, code review, documentation, refactoring, dependency updates, or single-file changes.

70

Quality

85%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

78%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

A well-structured, actionable research workflow with clear sequencing, explicit stop-condition checkpoints, and clean progressive disclosure via real reference files. Slight room to tighten repeated stop-rule prose and make metric extraction concrete.

Suggestions

Consolidate the stop-rule mechanics (currently spread across Loop steps 5/11, the Stop section, and the Avoid section) into one place to reduce repetition.

Add a concrete example command for extracting the primary metric from a run.log instead of leaving it as 'a command appropriate for the actual log format'.

Make the error-recovery feedback loop explicit (e.g. 'If the run crashes, inspect the last log lines, fix, and re-launch on the same branch with a follow-up commit') to raise workflow clarity toward 5.

DimensionReasoningScore

Conciseness

Mostly lean and assumes Claude's competence (e.g. minimal scaffolding around the bash snippet), but the stop-rule explanation repeats mechanics across several sections and could be tightened.

4 / 5

Actionability

Provides executable commands like `uv run <entrypoint> > "$LOG_DIR/run.log" 2>&1` and a concrete LOG_DIR pattern, but metric extraction is left generic ('a command appropriate for the actual log format'), a minor gap.

4 / 5

Workflow Clarity

Clear numbered sequence with explicit stop-condition re-checks (steps 5 and 11) and status marking; validation is present rather than absent, so the destructive/batch cap at 3 does not apply, though error-recovery feedback loops are only implicit.

4 / 5

Progressive Disclosure

SKILL.md is a concise overview with three clearly-signaled one-level-deep references (git-workflow.md, exploration-ideas.md, experiment-log-template.md), all verified to exist in references/.

5 / 5

Total

17

/

20

Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A high-quality description: concrete capabilities, explicit when-to-use triggers, and a strong negative boundary that distinguishes it from adjacent skills. Only slightly short of perfect trigger-term coverage (no file extensions or synonyms).

DimensionReasoningScore

Specificity

Lists multiple concrete actions — 'wiring RL or NeMo-gym runs, launching reproducible baselines and iterations, analyzing results, preserving human oversight, and using git plus TSV logs' — with comprehensive lifecycle coverage.

5 / 5

Completeness

Explicitly answers what (full experiment lifecycle) and when (when_to_use triggers plus an explicit 'Do NOT use for' boundary), matching the concrete-trigger-phrases anchor.

5 / 5

Trigger Term Quality

Natural user phrases like 'run experiments', 'find a better recipe', 'improve accuracy' map well, but coverage lacks file extensions and a few common synonyms, leaving minor gaps vs the 5 anchor.

4 / 5

Distinctiveness Conflict Risk

The negative boundary ('Do NOT use for: bug fixes, code review, documentation, refactoring, dependency updates, or single-file changes') carves a clear niche with minimal conflict risk.

5 / 5

Total

19

/

20

Passed

Validation

100%

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository
NVIDIA/skills
Reviewed

Table of Contents

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.