CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

setup

Guided setup for the vSR plugin. Runs the vLLM Semantic Router classifier on this machine's GPU or CPU or connects a remote runtime, picks which of vSR's six apps run, sets thresholds, writes the settings, and checks each app against the runtime. Use when vSR was just installed, when its tally says classifier offline or its rows say SKIPPED, or when the user wants to change vSR's runtime, apps, or thresholds.

Quality

89%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide
SecuritybySnyk

Passed

No findings from the security scan

SKILL.md
Quality
Evals
Security

Quality

Content

77%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.

An unusually rigorous procedural skill: every step has an explicit completion criterion, the settings write is guarded by a plan/diff/confirm/apply loop with backup, and error-recovery branches are enumerated down to specific HTTP codes. The two real weaknesses are references to local-runtime.md and thresholds.md, which are missing from the bundle, and dense run-on passages in steps 5 and 8 that could be tightened.

Suggestions

Add the missing local-runtime.md and thresholds.md to the bundle (e.g., under references/) or inline their essential content — the Local GPU/CPU branch in step 5 and the Advanced thresholds path in step 7 currently dead-end with no file to read.

Show the literal --set option names for the six app switches and the thresholds in step 8 (e.g., the exact NAME=VALUE forms settings.py accepts), so the write command can be built without guessing.

Break step 5's single long paragraph into a short bulleted structure per branch (Keep / Local / Remote) to trim the densest passage and make the warning-handling rules easier to follow.

DimensionReasoningScore

Conciseness

The body assumes Claude's competence — no basic concepts are explained — and nearly every sentence carries operational detail (exact commands, expected outputs, branch conditions), and each step's "Done when..." criterion is maximally terse. It is not a 5 because step 5 is a dense run-on paragraph ("the runtime line names the device it runs on when the runtime reports one and how long one question takes") and step 8's default-clearing semantics could be tightened without loss.

4 / 5

Actionability

Commands are copy-paste ready with full paths and flags ("python3 ${CLAUDE_SKILL_DIR}/scripts/settings.py plan --id <id> --set ...", "runtime.py health --url <url>") and expected outputs are specified ("prints `vsr@<marketplace>`", "prints `ready`", "`no change`", "`wrote`") plus error handling for HTTP 401/403/400 and "warning: slow runtime". It falls short of 5 because the Local GPU/CPU branch delegates entirely to local-runtime.md and Advanced thresholds to thresholds.md, neither of which exists in the bundle, and step 8 never shows the literal --set option names for the six app switches.

4 / 5

Workflow Clarity

Ten numbered steps each end with an explicit "Done when..." checkpoint; step 8 validates before writing (plan prints a diff, user confirms, then apply, with a backup kept); step 9 verifies every enabled app and explains "drift, flip or error"; and step 5 specifies recovery loops (rerun health and probe with --token-file, ask whether to keep a slow runtime or go back to step 4). This is the anchor case of explicit validation steps, feedback loops, and checkpoints, not the 4 with minor validation gaps.

5 / 5

Progressive Disclosure

The body itself is well structured — a ~56-line overview with clear step sections and clearly signaled links ("read [local-runtime.md](local-runtime.md) and follow it", "read [thresholds.md](thresholds.md)") — but neither referenced file exists in the bundle: there is no references/ directory and no local-runtime.md or thresholds.md at the skill root. Scoring against the actual bundle per the guidelines, the promised one-level-deep split is not realized, so it cannot score 4 ("references mostly clear; minor organization gaps") — two navigation paths dead-end.

3 / 5

Total

16

/

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.

A model description: third-person voice, a comprehensive list of concrete actions spanning the entire setup workflow, and an explicit 'Use when' clause whose triggers are observable plugin states a user would naturally repeat. It is both fully specific and fully distinct from any neighboring skill.

DimensionReasoningScore

Specificity

The description lists six concrete actions covering the whole setup flow: "Runs the vLLM Semantic Router classifier on this machine's GPU or CPU or connects a remote runtime", "picks which of vSR's six apps run", "sets thresholds", "writes the settings", and "checks each app against the runtime". This matches the anchor for multiple specific concrete actions with comprehensive coverage; nothing in the workflow is left unnamed, so it is not a 4.

5 / 5

Completeness

Both halves are explicit: the "what" enumerates the concrete actions (run/connect the classifier, pick apps, set thresholds, write settings, verify apps) and the "when" is a dedicated "Use when" clause with three concrete trigger conditions. This is the anchor example of clearly answering both what AND when with concrete trigger phrases, not the 4 where the when could be more specific.

5 / 5

Trigger Term Quality

The trigger phrases are exactly what a user of this plugin would say: "vSR was just installed", "its tally says classifier offline or its rows say SKIPPED", and "change vSR's runtime, apps, or thresholds", plus "Guided setup" covering setup/configure phrasing. Coverage of natural terms for this domain is comprehensive with plugin-observable states users would repeat verbatim; no common variation is missing, so it fits the top anchor rather than 4.

5 / 5

Distinctiveness Conflict Risk

The description is tightly scoped to one plugin — "vSR" appears throughout, plus plugin-specific surface terms like "vLLM Semantic Router", "tally", and "rows say SKIPPED" that no other skill would claim. It carves a clear niche with distinct triggers and minimal conflict risk, matching the top anchor; the 4 anchor's "minor overlap risk" does not apply.

5 / 5

Total

20

/

20

Passed

Validation

87%

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

CriteriaDescriptionResult

allowed_tools_field

'allowed-tools' contains unusual tool name(s)

Warning

relative_links

Relative link issues: 2 missing

Warning

Total

14

/

16

Passed

Repository

vllm-project/semantic-router

Path

src/agent-harness/claude-code/skills/setup/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.