SKILLmama scores libraries across five tiers to end stack-guessing
Magithar Sridhar built SKILLmama, an AI-native capability discovery engine that scans a project's architecture, searches five ecosystem tiers, and returns ranked library recommendations scored on compatibility, popularity, maintenance, and simplicity.
Score breakdown
SKILLmama replaces ad-hoc library selection with a transparent, multi-signal scoring system that explicitly surfaces MCP ecosystem options alongside traditional package registries.
- 01SKILLmama is an AI-native capability discovery engine built by Magithar Sridhar.
- 02Scoring formula: Compatibility (40%) + Popularity (30%) + Maintenance (15%) + Simplicity (15%), each dimension rated 1–10.
- 03Compatibility is weighted highest because a popular library with no client for your language is useless.
Magithar Sridhar built SKILLmama to address a common developer pain point: spending 45 minutes comparing libraries based on star counts and vibes, only to discover months later that the chosen package is unmaintained. The engine takes a capability gap as input, scans the project's architecture (reading files like `package.json`, Dockerfiles, and READMEs), and then runs a structured five-tier search to gather 8 or more candidates before scoring them.
The scoring formula is explicit: `Score = (Compatibility × 0.40) + (Popularity × 0.30) + (Maintenance × 0.15) + (Simplicity × 0.15)`.
The scoring formula is explicit: `Score = (Compatibility × 0.40) + (Popularity × 0.30) + (Maintenance × 0.15) + (Simplicity × 0.15)`. Compatibility is weighted highest because a library with 50,000 GitHub stars is useless if it lacks a client for the project's language. Popularity captures ecosystem health via GitHub stars and weekly download counts from npm or PyPI. Maintenance measures days since last commit and release cadence, while simplicity covers setup effort and documentation quality.
The five search tiers proceed in order — skills.sh, GitHub, the Smithery/MCP ecosystem, package registries (npm/PyPI/pkg.go.dev), and curated templates such as LangGraph and OpenHands cookbooks — stopping once 8 or more candidates are found. The post highlights Tier 3 (the MCP ecosystem) as particularly notable, since an available MCP server for a given capability could allow direct plug-in to an AI workflow instead of writing custom integration code. SKILLmama is described as compatible with Claude Code, Claude.ai, OpenAI Codex, and Antigravity.
Key facts
- 01SKILLmama is an AI-native capability discovery engine built by Magithar Sridhar.
- 02Scoring formula: Compatibility (40%) + Popularity (30%) + Maintenance (15%) + Simplicity (15%), each dimension rated 1–10.
- 03Compatibility is weighted highest because a popular library with no client for your language is useless.
- 04The engine searches five tiers in order: skills.sh, GitHub, MCP ecosystem (Smithery), npm/PyPI/pkg.go.dev, and curated templates.
- 05Search stops once 8 or more candidates are found across the tiers.
- 06Tier 3 surfaces MCP servers as a potential plug-in alternative to writing custom integration code.
- 07Compatible with Claude Code, Claude.ai, OpenAI Codex, and Antigravity.
Topics
Summary and scoring are generated automatically from the original article. We always link back to the publisher and never republish images or paywalled content. Last processed Jun 19, 2026 · 10:25 UTC. How this works →