CC deepspeed
Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention
Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention
As a process C 54/100 · Has gaps — weak spots: result and completion, execution cost, running it twice
How to improve
- For Hermes the description must be one sentence under 60 characters; move the conditions to a "When to Use" section.
- The SKILL.md body is over 5,000 tokens: move reference detail into references/ and load it when needed.
- Your own cases (evals/evals.json, 4–6 real requests with expected answers): the full check would then run those instead of a model-drafted suite.
- A spec.yaml with trigger phrases and assertions — a behaviour contract for CI; `skilltest init` writes a template.
Guard findings · 1
✓ No critical or high findings
Medium and low: 1
-
low Secrets in code
secret-high-entropy-tokenSKILL.md:67High-entropy token-like string (may be an id, hash or a credential) (quoted — discussed, not commanded)**Pattern 6:** Flops Profiler Contents Overview Flops Measurement Multi-GPU, Multi-node, Data Parallelism, and Model Parallelism Usage Usage With the DeepSpeed Runtime Example: Megatron-LM Usage Outsi
quoted
Files scanned: 9. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
- warning
description-long-hermesdescription is 149 chars; the Hermes authoring standard requires ≤ 60 (one sentence, ending with a period) - warning
body-longSKILL.md body ≈ 36084 tokens (recommended < 5000); move details to references/ - note
frontmatter-keyunknown frontmatter key "dependencies" - note
edit-residuethe text marks something as outdated (lines 43, 55): check that old rules are not kept next to new ones — the full check reads the text for contradictions
Process rating: all ten parameters 54/100
- 0Result and completion. Does not say what the result is
- 10Execution cost. Instruction body is 36084 tokens: crowds the task out of the window
- 30Running it twice. 42 mutating operations with no state check
- 60Tools and files. Uses tools (bash, web, python, node) that frontmatter does not declare
- 60Steps. 24 steps, 6 vague phrases
- 60Failures and branches. 2 branches
- 70When it triggers. States when to use, but not when not to
- 70Inputs and preconditions. Inputs and preconditions are listed
- 100Consistency. Name and required fields are in place
- 100Progress reporting. Reports progress
- low The response is described with custom markup (4 tags): a typed call is more reliable
Everything here is measured from the skill text rather than judged by a model, so the numbers are checkable. A parameter weighs more when it is a more common reason for the process to stall.
Quality signals
- +5Description has no quoted example phrases that should trigger the skill
- +4Description does not say when NOT to use the skill (false activations)
- +3Output format is not stated: the model decides each time
- -5TODO / placeholder text left in the skill
- +2Single-language instructions
- +3Description length 149: enough signal without eating the budget
- +4Structure: 17 headings
- +3Step-by-step instructions: 24 items
- +4Has examples (9 code blocks)
- +4Reference files are cited in the instructions (7 of 8)
- +1License stated
Quality base 70; lint remarks subtract, signals add up to 100. Result: 67.