feat(learning): add learning runtime with pattern mining

Ultraworked with [Sisyphus](https://github.com/code-yeongyu/oh-my-openagent)

Co-authored-by: Sisyphus <clio-agent@sisyphuslabs.ai>

backend/app/agents/learning/skill_candidate_builder.py

@@ -0,0 +1,54 @@
+from __future__ import annotations
+
+import hashlib
+
+from app.agents.schemas.learning import PatternCandidate, SkillCandidate
+
+
+class SkillCandidateBuilder:
+    def build(self, patterns: list[PatternCandidate]) -> list[SkillCandidate]:
+        candidates: list[SkillCandidate] = []
+
+        for pattern in patterns:
+            if pattern.confidence < 0.55:
+                continue
+
+            name = self._build_name(pattern)
+            candidates.append(
+                SkillCandidate(
+                    candidate_id=f"candidate-{self._stable_suffix(pattern)}",
+                    name=name,
+                    summary=pattern.description,
+                    candidate_type=self._map_candidate_type(pattern.pattern_type),
+                    source_pattern_ids=[pattern.pattern_id],
+                    confidence=pattern.confidence,
+                    evidence_refs=pattern.evidence_refs[:4],
+                    recommended_status="candidate",
+                )
+            )
+
+        return candidates
+
+    @staticmethod
+    def _build_name(pattern: PatternCandidate) -> str:
+        prefix = {
+            "workflow": "workflow",
+            "decomposition": "decomposition",
+            "preference": "preference",
+        }.get(pattern.pattern_type, "learned")
+        stable_suffix = SkillCandidateBuilder._stable_suffix(pattern)
+        return f"{prefix}-{stable_suffix}"
+
+    @staticmethod
+    def _map_candidate_type(pattern_type: str) -> str:
+        mapping = {
+            "workflow": "workflow_skill",
+            "decomposition": "decomposition_skill",
+            "preference": "preference_skill",
+        }
+        return mapping.get(pattern_type, "workflow_skill")
+
+    @staticmethod
+    def _stable_suffix(pattern: PatternCandidate) -> str:
+        raw = f"{pattern.pattern_type}:{pattern.description}".encode("utf-8")
+        return hashlib.sha1(raw).hexdigest()[:10]