CtrlK
BlogDocsLog inGet started
Tessl Logo

ai-sdk

Answer questions about the AI SDK and help build AI-powered features. Use when developers: (1) Ask about AI SDK functions like generateText, streamText, ToolLoopAgent, embed, or tools, (2) Want to build AI agents, chatbots, RAG systems, or text generation features, (3) Have questions about AI providers (OpenAI, Anthropic, Google, etc.), streaming, tool calling, structured output, or embeddings, (4) Use React hooks like useChat or useCompletion. Triggers on: "AI SDK", "Vercel AI SDK", "generateText", "streamText", "add AI to my app", "build an agent", "tool calling", "structured output", "useChat".

87

1.35x
Quality

83%

Does it follow best practices?

Impact

92%

1.35x

Average score across 3 eval scenarios

SecuritybySnyk

Passed

No findings from the security scan

The canonical home for this skill is ai-sdk in vercel/ai

SKILL.md
Quality
Evals
Security

Quality

Content

75%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.

A well-built, anti-drift skill: it consistently redirects from stale model memory to version-matched bundled docs, with concrete commands and paths for most operations. Main gaps are unnamed reference targets in two sections, mild repetition of the core warning, and the absence of an explicit recovery loop for type-check failures.

Suggestions

Name the actual file paths for the two unnamed references — replace 'read the bundled guide' and 'read the bundled DevTools documentation' with the specific paths under node_modules/ai/docs/ (and the DevTools package docs), matching the precision used elsewhere in the skill.

Consolidate the don't-trust-memory warning into a single section and reference it from 'Choosing a Model' and 'After Making Changes' instead of restating it, saving tokens without losing the point.

Add an explicit error-recovery sequence in 'After Making Changes' (type check fails → re-grep bundled docs for the changed API → fix → re-run), turning the current implicit loop into a checkpointed workflow.

DimensionReasoningScore

Conciseness

The body is dense with directives and avoids explaining background concepts, but it is mildly repetitive: the don't-trust-memory point is made three times ('Whatever you remember... is likely outdated', 'Never write AI SDK code from memory', and again in 'Choosing a Model' and 'After Making Changes'), and the 'What the AI SDK Is' paragraph is partially redundant with the description. Fits anchor 4 — efficient with minor instances that could be trimmed — rather than 5.

4 / 5

Actionability

Mostly executable guidance: concrete paths (node_modules/ai/docs/, <package>/node_modules/@ai-sdk/<name>/docs/), runnable commands (curl + jq model listing, npm view ai version, pnpm add ai --filter <pkg>), URL patterns (.md suffix, search-docs endpoint), and an env var name. It falls short of 5 because two sections defer with unnamed targets — 'read the bundled guide' and 'read the bundled DevTools documentation' give no file path, leaving the reader to discover what to open.

4 / 5

Workflow Clarity

The doc-verification workflow is a clear numbered sequence (check install → read bundled docs → fallback to web search → explicitly say so if unsupported), with checkpoints for version currency (installed vs npm view ai version) and post-change validation (run the type checker, re-check docs on errors). No destructive or batch operations, so no cap applies. Not 5 because there is no explicit fix-and-retry loop for, e.g., type-check failures — it says to re-check docs but doesn't sequence the recovery.

4 / 5

Progressive Disclosure

No bundle files exist (no references/, scripts/, or assets/), and the skill appropriately avoids inlining SDK documentation by delegating to the version-matched bundled docs at concrete one-level-deep paths, with well-organized sections. It falls short of 5 because 'the bundled guide' and 'the bundled DevTools documentation' are referenced without naming the specific files, creating minor navigation gaps.

4 / 5

Total

16

/

20

Passed

Description

91%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 strong description: explicit what/when structure, a rich explicit trigger list, and concrete API names throughout. The only weaknesses are the slightly generic opening verbs and a few overly broad trigger phrases that create minor conflict risk with adjacent LLM-tooling skills.

DimensionReasoningScore

Specificity

The 'what' verbs are somewhat generic ('Answer questions about the AI SDK and help build AI-powered features'), but the description enumerates concrete API names (generateText, streamText, ToolLoopAgent, embed, useChat, useCompletion) and capability areas (streaming, tool calling, structured output, embeddings). This matches anchor 4 — several specific actions with minor gaps — rather than anchor 5, whose 'what' statement itself is a comprehensive list of concrete actions.

4 / 5

Completeness

It explicitly answers 'what' ('Answer questions about the AI SDK and help build AI-powered features') and 'when' with a four-part 'Use when developers:' enumeration plus a dedicated 'Triggers on:' phrase list — matching the anchor-5 example's structure of concrete what + explicit trigger phrases. Not 4 because the when-clause is fully explicit, not merely present.

5 / 5

Trigger Term Quality

The explicit trigger list combines natural user phrasing ('add AI to my app', 'build an agent') with the exact technical terms developers would say ('AI SDK', 'Vercel AI SDK', 'generateText', 'streamText', 'tool calling', 'structured output', 'useChat'), plus synonyms spread across the when-clauses (chatbots, RAG, embeddings, providers). Coverage is comprehensive with both casual and technical variants.

5 / 5

Distinctiveness Conflict Risk

The skill has a clear niche (the Vercel AI SDK / `ai` package) and mostly distinct triggers, but broad phrases like 'build an agent', 'tool calling', and 'add AI to my app' could also fire for general LLM or other-provider work where the AI SDK isn't involved. Minor overlap risk with closely related skills fits anchor 4 rather than 5.

4 / 5

Total

18

/

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
vercel/vercel-plugin
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.