slime-rl-training

Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework. Use when training GLM models, implementing custom data generation workflows, or needing tight Megatron-LM integration for RL scaling.

Quality

72%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/coding/slime/SKILL.md

Quality

Content

61%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.

Actionable and reasonably concise content with well-sequenced workflows, but it underuses progressive disclosure by inlining reference material that duplicated bundle files already provide, and long-running training workflows lack explicit validation feedback loops.

Suggestions

Reference the existing bundle files from the body — e.g. replace the inlined 'Common Issues and Solutions' and 'Configuration Reference' sections with pointers to references/troubleshooting.md and references/api-reference.md — so SKILL.md stays an overview.

Add explicit validation feedback loops to the training workflows (e.g. check reward curves / checkpoint integrity, then fix and re-run) rather than a passive monitoring checklist, since RL training is a long-running batch operation.

Tighten the cost-warning, alternative-framework, and ASCII architecture sections to lean instruction, or move them to a reference file, to improve token efficiency.

Dimension	Reasoning	Score
Conciseness	Largely lean with copy-paste commands and minimal over-explanation of known concepts, though contextual padding (cost warnings, alternative-framework comparisons, ASCII architecture diagram) could be trimmed.	4 / 5
Actionability	Highly executable with concrete docker/python commands, real flags, and JSONL/Python examples, but a few examples use placeholders ('/path/to/data.jsonl') or undefined helpers ('compute_reward').	4 / 5
Workflow Clarity	Three workflows have sequenced numbered steps and prerequisite checklists, but long-running/batch RL training lacks explicit validate->fix->retry feedback loops, capping this dimension at 3 per the rubric.	3 / 5
Progressive Disclosure	The body is a monolithic ~466-line file that inlines config-reference and troubleshooting content despite bundle files references/api-reference.md and references/troubleshooting.md existing but never being referenced from the body.	2 / 5
	Total	13 / 20 Passed

Description

83%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 strong, specific description that clearly states both what the skill does and when to use it, with concrete trigger terms and a distinct niche. Minor room to add synonymous trigger phrasings and enumerate capabilities more directly.

Dimension	Reasoning	Score
Specificity	Names the domain ('LLM post-training with RL using slime, a Megatron+SGLang framework') and lists several concrete use actions ('training GLM models', 'implementing custom data generation workflows', 'needing tight Megatron-LM integration'), with minor gaps in capability coverage.	4 / 5
Completeness	Explicitly answers both what ('Provides guidance for LLM post-training with RL using slime, a Megatron+SGLang framework') and when ('Use when training GLM models, implementing custom data generation workflows, or needing tight Megatron-LM integration for RL scaling') with concrete trigger phrases.	5 / 5
Trigger Term Quality	Good natural keywords users would say ('training GLM models', 'custom data generation workflows', 'Megatron-LM integration', 'RL scaling'), though a few common synonyms (e.g. RLHF, GRPO, post-training) are missing.	4 / 5
Distinctiveness Conflict Risk	Has a clear niche (slime / Megatron+SGLang / GLM RL post-training) with distinct triggers, with minor overlap risk against general RL post-training skills like verl or miles.	4 / 5
	Total	17 / 20 Passed

Validation

87%

Warnings & errors only

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

Validation — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 4 missing, 2 deeper-than-1-level	Warning

	Total	14 / 16 Passed

Repository: synthetic-sciences/openscience
Path: backend/cli/skills/coding/slime/SKILL.md
Commit: 3a6c3a9

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

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.