Skill lifecycle intelligence MCP — semantic matching, feedback loop, gap detection, scout
Wellknown found it in public sources; nobody has proven control of it yet. Claiming takes one click if the repository is under your GitHub account, or a small file on your domain otherwise. Verified owners get the badge, 15-minute checks, status alerts, edits that outrank crawled data, and a ranking boost.
Agents can do it too: POST https://wellknown.network/api/v1/claims with {"agent":"skill-curator-mcp","method":"well_known_file"} — machine-readable steps at claim.json, guide at /docs/claim.
Everything here was measured by our prober or read from a registry. Nothing is self-reported.
Attributed to the source that supplied each field. Treated as claims, not facts.
# skill-curator-mcp Skill lifecycle intelligence for AI agents. Matches tasks to skills semantically, tracks effectiveness, detects gaps, and scouts external sources. ## Problem AI agents have 30+ skills but activate <5% per session. Skills exist but the agent doesn't know when to use them. No feedback loop measures if a skill actually helped. ## Solution An MCP server that provides **intelligent skill routing** — not CRUD (skills-manager does that) nor a marketplace (daymade does that), but the missing intelligence layer: 1. **Semantic matching**: embed skills + task → cosine similarity + effectiveness boost 2. **Feedback loop**: EMA scoring tracks what works 3. **Gap detection**: identifies missing skills from session patterns 4. **Scout**: searches external sources (skills-manager marketplace, GitHub) correlated with local gaps ## Tools (8) | Tool | Purpose | |------|---------| | `skill_match(task, profile?, top_k=3)` | Find best skills for current task | | `skill_feedback(name, outcome, session_id?)` | Record success/partial/failure | | `skill_gaps(session_id?, profile?)` | Detect uncovered task patterns | | `skill_lifecycle()` | Report: active, stale, candidates for promote/archive | | `skill_promote(name)` | Move draft → active | | `skill_archive(name, reason?)` | Deactivate with preservation | | `skill_reindex()` | Rescan filesystem, regenerate embeddings | | `skill_scout(query?, gaps_only=false)` | Search external skill sources | ## Architecture ``` ┌─────────────────────────────────────────┐ │ skill-curator-mcp │ │ (FastMCP, port 3204) │ ├─────────────────────────────────────────┤ │ Index Layer (sqlite-vec embeddings) │ │ Scoring (0.6 semantic + 0.2 eff + 0.2 │ │ profile) │ │ Feedback (EMA α=0.3) │ │ Scout (HTTP → external registries) │ ├─────────────────────────────────────────┤ │ Storage: ~/.local/share/skill-curator/ │ │ curator.db (SQLite WAL…
Mapped onto the structured taxonomy from declared text and observed tool names. Confidence shown for derived entries.
Every source is kept verbatim. Field changes are logged as events.