Tessl • The package manager for agent skills

SkillPublic

benchmark-sandbox

Run vercel-plugin eval scenarios in Vercel Sandboxes instead of local WezTerm panels. Provisions ephemeral microVMs with Claude Code + plugin pre-installed, runs benchmark prompts, extracts hook artifacts, and produces coverage reports.

1.38x

Quality

78%

Does it follow best practices?

Impact

98%

1.38x

2 of 3 eval scenarios. Add 1 more for a full score.

Securityby

Passed

No findings from the security scan

Fix and improve this skill with Tessl

tessl review fix ./.claude/skills/benchmark-sandbox/SKILL.md

Quality

Content

70%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 impressively actionable, hard-won operational playbook with an exemplary validated 3-phase workflow, undermined by monolithic structure and duplication. The missing run-eval.ts bundle file is the biggest integrity gap, since every command depends on it.

Suggestions

Ship run-eval.ts in scripts/ (or split the pipeline into scripts/ + references/) so the referenced runner actually exists in the bundle.

Deduplicate: merge 'Commands' into 'Proven Working Script', collapse 'How It Works' and 'Sandbox Session Flow' into one section, and state each hard-won fact (snapshot stops sandbox, v2 beta 404, timestamped slugs) exactly once.

Move the three score schemas, the dated 'Proven Results (2026-03-10)' metrics, and the monitoring TypeScript snippets into references/ files linked one level deep from SKILL.md.

Dimension	Reasoning	Score
Conciseness	Mostly efficient, high-signal operational content, but with notable duplication: the CLI usage examples appear nearly verbatim in both 'Proven Working Script' and 'Commands', 'How It Works' and 'Sandbox Session Flow' describe the same pipeline twice, and snapshot/v2-beta/timestamp-slug facts each recur 2-3 times. The dated 'Proven Results (2026-03-10)' section is time-sensitive content outside any deprecated/old-patterns section.	3 / 5
Actionability	Concrete, executable guidance throughout — copy-paste bun commands, a complete CLI flag table, a scenario JSON example, and exact SDK calls (sandbox.writeFiles, extendTimeout). Falls short of 5 because the central run-eval.ts is referenced but not present in the bundle (no scripts/ directory), and the monitoring snippets are fragments rather than runnable programs.	4 / 5
Workflow Clarity	The 3-phase BUILD/VERIFY/DEPLOY pipeline is explicitly sequenced with per-phase timeouts, haiku structured scoring as validation checkpoints after each phase, feedback loops (deploy fix-and-retry up to 3 attempts, verify fix-re-verify), a snapshot safety net, and crash-safe incremental result writes — a textbook batch workflow with validation.	5 / 5
Progressive Disclosure	Section structure is clear, but the skill is a 388-line monolith: score schemas, dated results, monitoring code, and the comparison table are all inlined where they belong in separate reference files, and the only referenced file (run-eval.ts) does not exist in the bundle. Better than the headerless inline-wall anchor (2) but clearly short of well-signaled working references (4).	3 / 5
	Total	15 / 20 Passed

Description

75%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 specific, third-person description with concrete actions and excellent distinctiveness, held back only by the absence of any 'use when' trigger guidance and modest synonym coverage. It would reliably fire when explicitly named but may not be surfaced from paraphrased requests.

Suggestions

Append an explicit trigger clause, e.g. 'Use when the user asks to run, parallelize, or automate vercel-plugin benchmark/eval runs, or mentions sandboxes instead of local WezTerm panels.'

Add natural synonym terms users would say ('plugin testing', 'benchmark the plugin', 'eval run', 'skill coverage') to broaden trigger matching.

Dimension	Reasoning	Score
Specificity	Lists multiple specific concrete actions — 'Provisions ephemeral microVMs with Claude Code + plugin pre-installed, runs benchmark prompts, extracts hook artifacts, and produces coverage reports' — with comprehensive coverage of the skill's capabilities, matching the top anchor.	5 / 5
Completeness	The 'what' is clear and concrete, but there is no 'Use when...' clause or equivalent explicit trigger guidance; 'instead of local WezTerm panels' only weakly implies when. Per the judging guideline, a missing trigger clause caps completeness at 3.	3 / 5
Trigger Term Quality	Good natural keyword coverage ('eval scenarios', 'benchmark prompts', 'coverage reports', 'Vercel Sandboxes'), but a few natural terms are missing (e.g., 'plugin testing', 'hook coverage') and there is no synonym variation, so it sits between the 4 and 5 anchors, closer to 4.	4 / 5
Distinctiveness Conflict Risk	Highly niche triggers ('vercel-plugin eval scenarios', 'Vercel Sandboxes', 'WezTerm panels') make it clearly distinguishable from any other skill with minimal conflict risk, matching the 5 anchor.	5 / 5
	Total	17 / 20 Passed

Validation

100%

Warnings & errors only

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: vercel/vercel-plugin
Path: .claude/skills/benchmark-sandbox/SKILL.md
Revision: 0db4408

Reviewed: 14 days ago

Discovery Implementation Validation

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.

benchmark-sandbox

Table of Contents

Is this your skill?