vector-index-tuning
Optimize vector index performance for latency, recall, and memory. Use when tuning HNSW parameters, selecting quantization strategies, or scaling vector search infrastructure.
77
71%
Does it follow best practices?
Impact
84%
1.27xAverage score across 3 eval scenarios
Passed
No known issues
Optimize this skill with Tessl
npx tessl skill review --optimize ./plugins/llm-application-dev/skills/vector-index-tuning/SKILL.mdQuality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is an excellent skill description that clearly defines a specific technical niche. It uses precise domain terminology that practitioners would naturally use, provides both what the skill does and when to use it, and occupies a distinct space unlikely to overlap with other skills.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: optimizing vector index performance across three dimensions (latency, recall, memory), tuning HNSW parameters, selecting quantization strategies, and scaling vector search infrastructure. | 3 / 3 |
Completeness | Clearly answers both 'what' (optimize vector index performance for latency, recall, and memory) and 'when' (explicit 'Use when' clause covering tuning HNSW parameters, selecting quantization strategies, or scaling vector search infrastructure). | 3 / 3 |
Trigger Term Quality | Includes strong natural keywords users would say: 'vector index', 'HNSW', 'quantization', 'vector search', 'latency', 'recall', 'memory'. These are terms practitioners naturally use when dealing with vector search optimization. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive niche targeting vector index optimization specifically. Terms like 'HNSW parameters', 'quantization strategies', and 'vector search infrastructure' are very specific and unlikely to conflict with other skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
42%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
The skill provides genuinely useful, executable code for vector index tuning with good parameter tables and real library integrations. However, it is far too verbose—most of the code templates are standard patterns Claude could generate without instruction, and the content would benefit enormously from being split into a concise overview with references to separate template files. It also lacks an explicit end-to-end workflow with validation checkpoints.
Suggestions
Extract the four code templates into separate files (e.g., templates/hnsw_benchmark.py, templates/quantization.py) and replace them in SKILL.md with brief descriptions and links, keeping only the decision tables and parameter recommendations inline.
Add an explicit numbered workflow: 1) Select index type based on data size → 2) Benchmark HNSW parameters → 3) Validate recall meets target → 4) Apply quantization → 5) Re-validate recall → 6) Deploy and monitor. Include validation checkpoints between steps.
Remove code that Claude can trivially generate (e.g., the full VectorQuantizer class, binary packing logic, basic KMeans usage) and focus on the non-obvious configuration values, tradeoff guidance, and Qdrant-specific API patterns.
Add a concise 'Quick Reference' section at the top with just the parameter recommendation rules (e.g., 'For 1M vectors targeting 95% recall: M=32, efConstruction=200, efSearch=128, INT8 quantization') so Claude can act without reading the full document.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is extremely verbose at ~400+ lines, with massive code templates that Claude could generate on its own. The quantization class, monitoring code, and benchmarking functions are standard patterns that don't need to be spelled out in full. The reference tables (index selection, HNSW parameters, quantization types) are valuable but buried in excessive boilerplate. | 1 / 3 |
Actionability | The code templates are fully executable with real libraries (hnswlib, qdrant_client, sklearn), include concrete parameter values, and cover the full workflow from benchmarking to production configuration. The Qdrant configuration template with preset optimization profiles is immediately usable. | 3 / 3 |
Workflow Clarity | While individual templates are clear, there's no overarching workflow connecting them (e.g., 'first benchmark → then select quantization → then configure production → then monitor'). There are no validation checkpoints—for instance, no step to verify recall meets targets before deploying a configuration, or to validate that quantization hasn't degraded quality below acceptable thresholds. | 2 / 3 |
Progressive Disclosure | This is a monolithic wall of code with no references to external files. The four large templates should be split into separate reference files, with SKILL.md containing only the decision tables, parameter recommendations, and links. Everything is inline, making it hard to navigate. | 1 / 3 |
Total | 7 / 12 Passed |
Validation
90%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 10 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
skill_md_line_count | SKILL.md is long (518 lines); consider splitting into references/ and linking | Warning |
Total | 10 / 11 Passed | |
- Repository
- wshobson/agents
- Commit
91fe43e
- 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.