AC evalscope
LLM evaluation & inference performance testing via the evalscope CLI. Translates natural language requests into evalscope commands for: (1) Model accuracy evaluation — runs 160+ benchmarks against local checkpoints or API endpoints (OpenAI-compatible, Anthropic, LiteLLM); (2) Performance stress testing — TTFT, TPOT, throughput, latency under configurable concurrency; (3) RAG evaluation — RAGAS quality metrics, MTEB embedding benchmarks, CLIP retrieval; (4) Benchmark discovery — list/filter/inspect benchmarks by tag. Trigger on: evaluate / benchmark / score a model, throughput / latency / QPS / stress test, find benchmarks, view results, 评测模型, 压测, 跑 benchmark, 性能测试, 查看评测结果, 有哪些评测集, RAG 评测, embedding 评测. Do NOT trigger for: model training / finetuning / deployment / serving requests.
As a process C 62/100 · Has gaps — weak spots: result and completion, consistency
How to improve
- 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 · 0
✓ No critical or high findings
Files scanned: 7. Evidence is masked. Grey chips explain why severity was lowered.
Against the Agent Skills spec
✓ No remarks against the Agent Skills spec
Process rating: all ten parameters 62/100
- 0Result and completion. Does not say what the result is
- 40Consistency. Frontmatter name (evalscope) differs from the folder (skill-evalscope)
- 50Failures and branches. 0 branches, has a failure section
- 60Tools and files. Uses tools (python) that frontmatter does not declare
- 70When it triggers. States when to use, but not when not to
- 70Inputs and preconditions. Inputs and preconditions are listed
- 100Steps. 28 steps
- 100Execution cost. Instruction body is 1652 tokens
- 100Running it twice. No mutating operations
- 100Progress reporting. Reports progress
- low 10 top-level sections: this looks like several domains in one skill
- low The skill ranks results itself: that belongs to the system behind the tool, not the model
- 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
- -2localhost URLs: will not work for another user
- +1No license
- +2Single-language instructions
- +3Description length 792: enough signal without eating the budget
- +4Structure: 11 headings
- +3Step-by-step instructions: 28 items
- +4Has examples (7 code blocks)
Quality base 70; lint remarks subtract, signals add up to 100. Result: 82.