Conciseness

The body is compact and table-based, but the 'AI Search Capabilities' and 'Speech Capabilities' tables restate concepts Claude already knows ("Full-text search — Linguistic analysis, stemming", "Vector search — Semantic similarity with embeddings", "Text-to-speech — Neural voices, SSML support"), and service information is repeated across the Services table, MCP section, and capability tables (OCR appears three times). This matches anchor 3, 'mostly efficient but includes some unnecessary explanation or could be tightened', rather than anchor 4 since two full sections could be cut.

Actionability

Concrete, executable guidance is present: exact MCP tool and command names ("azure__search with command search_index_list", "azure__speech with command speech_synthesize"), CLI commands ("az search", "az cognitiveservices"), and an explicit MCP-not-enabled fallback ("ask the user to run /mcp and enable Azure MCP"). It falls short of anchor 5 because there are no example invocations with arguments (e.g., a sample search_query payload or index name) and no CLI path at all for Speech and Document Intelligence.

Workflow Clarity

As a routing/reference skill the single action is largely unambiguous — prefer Azure MCP tools, with a clearly stated fallback when MCP is absent — and the operations involved are read-only so no validation cap applies. It sits at anchor 4 rather than 5 because the fallback path ends at 'ask the user to enable' with no guidance for proceeding (e.g., when the user declines or auth fails, and the bundle's auth-best-practices.md is never linked).

Progressive Disclosure

The body is a clear overview with well-signaled, one-level-deep references: all 14 SDK links (e.g., references/sdk/azure-search-documents-py.md) resolve to real files in references/sdk/, plus two external documentation links. However, scoring against the actual bundle, references/auth-best-practices.md exists but is never referenced anywhere in SKILL.md, making it undiscoverable — a minor organization gap that holds this at anchor 4 rather than anchor 5.

Specificity

The description lists several concrete actions — "search, vector/hybrid search, speech-to-text, text-to-speech, transcription, OCR" — covering the named services. It sits at anchor 4 rather than 5 because the OpenAI and Document Intelligence services are named as domains only, with no accompanying actions ("Use for Azure AI: Search, Speech, OpenAI, Document Intelligence"), leaving minor coverage gaps.

Completeness

It explicitly answers both questions: the 'what' ("Use for Azure AI: Search, Speech, OpenAI, Document Intelligence. Helps with search, vector/hybrid search, speech-to-text, text-to-speech, transcription, OCR.") and the 'when' with concrete trigger phrases ("WHEN: AI Search, query search, vector search, hybrid search, semantic search, speech-to-text, text-to-speech, transcribe, OCR, convert text to speech."). This matches anchor 5; anchor 4 would apply only if the WHEN guidance were less explicit.

Trigger Term Quality

The WHEN clause includes strong natural trigger phrasing with synonyms ("speech-to-text", "transcribe", "text-to-speech", "convert text to speech", "semantic search", "OCR"), but a few natural terms users would say are missing (e.g., "embeddings", "form extraction", "image analysis"), matching anchor 4's 'good coverage, a few natural terms missing' rather than anchor 5's comprehensive coverage.

Distinctiveness Conflict Risk

"Azure AI" with service names (AI Search, Speech, Document Intelligence) carves a clear niche, but generic triggers like "search", "transcribe", and "OCR" without an Azure qualifier create minor overlap risk with closely related document/PDF or AI skills — anchor 4 rather than anchor 5's 'minimal conflict risk'.