Skillshare vs Voyage AI

Skillshare's feature set is about movement and safety of skills: syncing a single source directory into 65+ CLI tools via symlinks (NTFS junctions on Windows), auto-detecting installed CLIs like Claude Code, Cursor, Codex and Gemini CLI, and non-destructive merges that preserve CLI-specific skills. Bidirectional sync collects edits made in any target back to source; GitHub search installs remote skills. Its distinctive capability is security: audits scan for prompt injection, data exfiltration and destructive commands, with critical threats blocked automatically and a --force override on updates. The 0.21.4 release adds plugins and MCP servers with zero Agent targets (install later by ticking a target), fixes scoped npm plugin blocking for Pi and OpenCode, and switches config.yaml writes to 2-space indentation. Dual-level scoping covers org-wide repos plus project .skillshare/ directories.

Voyage AI is a retrieval-model vendor, not an agent tool. Its lineup spans general-purpose voyage-3.5 and voyage-3.5 lite, domain-specific models for finance, legal and code, and company-specific fine-tunes on proprietary data, plus the Voyage 4 series. voyage-multimodal-3.5 extends embeddings to images and text; voyage-context-3 preserves global document context alongside chunk detail. rerank-2.5 and rerank-2.5-lite add instruction following. The engineering pitch is cost: 3x-8x shorter vectors, 4x smaller models, 2x cheaper inference with superior accuracy, 32K-token context and a Batch API for large embedding jobs.

They share vocabulary ("context," "skills") but zero functional overlap.

Skillshare is free and open source — there's no tier to buy. Your real cost is operational: a source directory (e.g. ~/.config/skillshare/skills/), git for syncing across machines, and the discipline to keep skills in one place. There's no hosted plan; the tool itself handles symlink propagation and audits locally. The --force flag gates security-audit overrides, so there's no surprise billing, just an explicit decision. For a solo dev or an open-source maintainer this is effectively zero marginal cost per added CLI.

Voyage AI is pricing_type "contact" — no public rates in the data, which usually signals usage- or contract-based enterprise pricing around embedding and rerank calls. That shapes who can buy: it's structured for enterprise RAG teams with storage/search budgets, not for someone testing whether RAG works. The cost argument Voyage makes is downstream — shorter embeddings (3x-8x) shrink vector storage and search bills, and inference is described as 2x cheaper with a 4x smaller model. So the comparison isn't price vs price; it's "free tool you run yourself" vs "a sales conversation to get retrieval accuracy and bill reduction at enterprise scale." Regulated buyers note the SOC 2 and HIPAA framing, which typically accompanies a paid enterprise agreement rather than self-serve checkout.