CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

vllm-sr-agent-operations

Install, configure, verify, maintain, troubleshoot and optimize vLLM Semantic Router v0.5 on a Docker host (CPU, AMD or NVIDIA GPU) or a Kubernetes cluster (Helm or the operator), through its CLI and Router API. Use for a first install, configuration and routing changes, checks of routed requests, upgrades and the v0.5 config migration, secrets, logs, health, backup and restore, removal, failure diagnosis, and tuning of the decision model, recipes, caches, latency and cost with sr-bench.

Quality

100%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Low

Low-risk findings worth noting

SKILL.md
Quality
Evals
Security

Quality

Content

100%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.

The body is an exemplary operations router: preflight checks are executable and table-driven, each task names preconditions, a reference, exact success evidence, an on-failure path, and stop-and-ask conditions, and feedback loops (rerun-after-fix, one-change-at-a-time, backup/rollback) are built in. Bundle structure matches the progressive-disclosure ideal — a concise overview with well-signaled, one-level-deep references that all exist. The only minor nit is that reference links use absolute URLs rather than relative paths, which does not impair navigation given unambiguous file naming.

DimensionReasoningScore

Conciseness

The body is dense and operational throughout: ground rules are terse imperatives ('Nothing waits for a person', 'Keep secrets in environment variables, never in YAML'), preflight is a runnable command block plus a compact pass/otherwise table, and each task section is a tight Preconditions/Run/Success/On-failure/Stop-and-ask block. The opening paragraph is product-specific architecture (System One, ports), not concepts Claude already knows, and version mentions (v0.5) are operational facts used as success criteria rather than date-conditional guidance. Every token earns its place for a skill covering this much surface.

5 / 5

Actionability

Guidance is copy-paste ready with exact expected outputs: preflight commands, a /dev/tcp port scan, docker inspect format strings, and success criteria pinned to exact strings and exit codes ('"config_version": "v0.5"', '✓ vLLM Semantic Router is running', '"applied": true', '"routing_decision": "cs-route"', '"ready": true', '"status": "completed"'). The single elided placeholder ('vllm-sr route probe …') is delegated to the existing route-verification reference, and the common cases are covered with executable specifics, matching the fully-executable anchor.

5 / 5

Workflow Clarity

Sequences are explicit ('steps 1 to 5: choose the path, install the CLI, write and validate the configuration, start the stack, check it'), every task has a validation checkpoint in its Success field, and error-recovery feedback loops are stated ('Change one thing at a time, then rerun the step that failed'; 'run that command if these rules allow it, then rerun the step'; Troubleshoot: 'run its check, apply its fix, rerun the failed step'). Destructive operations are protected: backups are required before upgrade/migration with a rollback path, so the missing-validation cap does not apply.

5 / 5

Progressive Disclosure

Scored against the actual bundle: SKILL.md is a lean router file whose Tasks table maps each task (with a 'When' column) to one of nine reference files, all of which exist. References are one level deep; cross-links between references (e.g., troubleshooting to configuration-loop sections) are lateral and anchored navigation aids rather than buried content, and http-configuration-plan.md is reachable from configuration-loop.md. Content is appropriately split, with detailed step loops living in the references.

5 / 5

Total

20

/

20

Passed

Description

100%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.

An exemplary description: it states concrete capabilities, platforms, and interfaces in the first sentence, then provides an explicit and thorough 'Use for' trigger list covering the full operational lifecycle. It follows third-person/imperative voice and names a distinct product, so it is both highly discoverable and unlikely to fire for the wrong skill.

DimensionReasoningScore

Specificity

The description lists six concrete verbs ('Install, configure, verify, maintain, troubleshoot and optimize'), names the target platforms and interfaces ('Docker host (CPU, AMD or NVIDIA GPU) or a Kubernetes cluster (Helm or the operator), through its CLI and Router API'), and enumerates a comprehensive task inventory from 'first install' through 'removal' and 'tuning of the decision model, recipes, caches, latency and cost with sr-bench'. This matches the anchor for multiple specific concrete actions with comprehensive coverage, not the anchor-4 case of minor gaps.

5 / 5

Completeness

It explicitly answers both questions: the first sentence states what the skill does, and 'Use for a first install, configuration and routing changes, checks of routed requests, upgrades and the v0.5 config migration, secrets, logs, health, backup and restore, removal, failure diagnosis, and tuning…' gives concrete 'when' triggers. This mirrors the rubric's own anchor-5 example structure with even fuller enumeration.

5 / 5

Trigger Term Quality

Natural trigger terms a user would actually say are comprehensively covered: 'install', 'configure', 'upgrade', 'migrate', 'secrets', 'logs', 'health', 'backup and restore', 'removal', 'failure diagnosis', 'latency', 'cost', plus the product and platform names ('vLLM Semantic Router', 'Docker', 'Kubernetes', 'Helm', 'sr-bench'). Synonym variants are present (troubleshoot/failure diagnosis, removal, tuning/optimize), so the anchor-4 case of missing natural terms does not fit.

5 / 5

Distinctiveness Conflict Risk

It names a specific product and version ('vLLM Semantic Router v0.5') with niche-specific triggers (v0.5 config migration, decision model, recipes, sr-bench, Router API), giving it a clear niche with minimal conflict risk against other skills. Voice is imperative/third person, so no person-penalty applies.

5 / 5

Total

20

/

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.

Validation — 16 / 16 Passed

Validation for skill structure

No warnings or errors.

Repository

vllm-project/semantic-router

Path

tools/agent/skills/vllm-sr-agent-operations/SKILL.md

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.