CtrlK
BlogDocsLog inGet started
Tessl Logo
SkillPublic

multi-model-orchestrator

Orchestrate multi-stage AI pipelines that route tasks to optimal models by role — fast models for classification, reasoning models for planning, content models for generation. Use when building "multi-model pipelines", "AI orchestration", "model routing", or "LLM orchestration".

Quality

82%

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

71%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-sequenced, highly actionable orchestration guide: lean mostly-code sections, a clear 9-step pipeline with per-stage timeout/fallback tables, and practical troubleshooting. The main structural weakness is the absence of progressive disclosure — everything is inlined in one 350-line file with no references/ bundle — and a few stubbed code paths that keep it from copy-paste completeness.

Suggestions

Move the preset tables (Step 2), provider adapter implementations (Step 3), RAG integration (Step 6), and multi-agent analysis (Step 7) into references/ files (e.g. references/presets.md, references/providers.md), keeping SKILL.md as a lean overview with clearly signaled one-level-deep links.

Define or inline the undefined helpers (`generateImages`, `persistResult`, `callGemini`, `classificationPrompt`, `planningPrompt`) and replace the '// Similar implementations for callCerebras, callOpenAI...' stub with real implementations or an explicit pointer to the referenced file.

Add an explicit validation checkpoint in the pipeline itself — wrap `JSON.parse` of classification/planning output in a validate-and-retry-once loop rather than relegating that guidance to the Troubleshooting table.

DimensionReasoningScore

Conciseness

The body is dense — mostly executable TypeScript, tables, and terse rules ('Use `Promise.allSettled` when partial failure is acceptable'), assuming Claude's competence with no basic-concept explanations. Minor trims possible: the 'Quick Reference' table and the 'Architects multi-stage AI pipelines...' paragraph largely restate the frontmatter description. Not 5 for that duplication; not 3 because there is no real padding beyond it.

4 / 5

Actionability

Concrete, runnable-style TypeScript for ModelFactory, concurrency limiting, SSE streaming, RAG retrieval, and fallback generation, plus a specific timeout/fallback table. Not 5 because key pieces are stubbed or undefined in-file: '// Similar implementations for callCerebras, callOpenAI...', and helpers like `generateImages`, `persistResult`, `callGemini`, `classificationPrompt`, `planningPrompt` are referenced but never defined; not 3 because what is written is executable, not pseudocode.

4 / 5

Workflow Clarity

A clear 9-step sequence from stage definition through error handling, with per-stage timeouts, on-timeout behavior, preset fallbacks, and a troubleshooting table with fixes. Not 5 because the pipeline code does raw `JSON.parse(...)` on model output with no inline validation/retry loop — the 'validate and retry once' guidance lives only in Troubleshooting; not 3 because the sequence and error-recovery paths are otherwise explicit.

4 / 5

Progressive Disclosure

Good section structure and a Cross-References list, but the skill is a single ~350-line monolithic file with no bundle files at all (no references/, scripts/, or assets/ exist). Content that belongs in separate files is inlined — preset tables, provider adapter implementations, RAG integration, and multi-agent analysis each warrant a references/ file — and the Cross-References point to sibling skills rather than navigable bundle paths. Not 2 because sections are clearly headed and navigable; not 4 because nothing is actually split out.

3 / 5

Total

15

/

20

Passed

Description

83%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: third-person voice, concrete role-to-model mapping, and an explicit 'Use when' clause with natural quoted trigger phrases including synonyms. The only weaknesses are minor coverage gaps (images, RAG, fallbacks) and slightly broad orchestration triggers that could overlap with adjacent skills.

DimensionReasoningScore

Specificity

Quotes: 'route tasks to optimal models by role — fast models for classification, reasoning models for planning, content models for generation' — several concrete, specific actions tied to role-based routing. Falls short of 5 because coverage has minor gaps (image generation, RAG retrieval, and fallback routing that the body documents are absent); above 3 because it lists multiple concrete actions rather than 1-2.

4 / 5

Completeness

Explicitly answers what ('Orchestrate multi-stage AI pipelines that route tasks to optimal models by role...') and when ('Use when building "multi-model pipelines", "AI orchestration"...') with concrete trigger phrases — a direct match for the anchor-5 good example.

5 / 5

Trigger Term Quality

Triggers: '"multi-model pipelines", "AI orchestration", "model routing", or "LLM orchestration"' — natural phrases a user would say, including the AI/LLM synonym pair. Not 5 because common variations like 'model selection', 'prompt routing', or 'cheap vs expensive model' are missing; not 3 because coverage is good with synonyms present.

4 / 5

Distinctiveness Conflict Risk

Clear multi-model routing niche with distinct triggers, so minimal conflict risk overall. Not 5 because 'AI orchestration' and 'LLM orchestration' are somewhat broad and could overlap with generic agent-orchestration or workflow skills; not 3 because the multi-model routing framing is well differentiated.

4 / 5

Total

17

/

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

paolomoz/skills

Path

skills/multi-model-orchestrator/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.