Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Bridge Claude Code auto-memory into AgentDB with ONNX embeddings, deduplicate, and enable unified cross-project search
.claude/skills/ruvnet-memory-bridge/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-05 | ✗→✓ | ▲ Improved | -61% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -61% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -75% | 0% |
| case-10 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-14 | ✗→✓ | ▲ Improved | -47% | 0% |
Import Claude Code's native auto-memory files into AgentDB for semantic search across sessions and projects.
Claude Code stores memories as markdown files in ~/.claude/projects/*/memory/*.md. This bridge:
claude-memories namespace with HNSW indexingmcp__plugin_ruflo-core_ruflo__memory_bridge_status({}) Verify: Claude files count, AgentDB entries, SONA state, connection status.
mcp__plugin_ruflo-core_ruflo__memory_import_claude({})mcp__plugin_ruflo-core_ruflo__memory_import_claude({ allProjects: true })CLI alternative: bash node .claude/helpers/auto-memory-hook.mjs import-all
mcp__plugin_ruflo-core_ruflo__memory_bridge_status({}) Confirm entry counts match expected file counts.
Search for near-duplicate entries (cosine > 0.95) and merge them, keeping the most recent version.
mcp__plugin_ruflo-core_ruflo__memory_search_unified({ query: "test query", limit: 3 }) Results include source attribution: claude-code, auto-memory, or agentdb.
The bridge runs automatically on session-start via the SessionStart hook. Manual invocation is only needed for:
When ruflo-ruvector is loaded, bridged memories are also indexed by ruvector for:
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 7,077 | 1,777 | -75% | 1 | 1 | 0% | 512 | 844 | +65% | 0 | 0 | — |
case-02 | fail→fail | 7,195 | 7,554 | +5% | 1 | 1 | 0% | 399 | 749 | +88% | 0 | 0 | — |
case-03 | fail→fail | 9,401 | 3,832 | -59% | 1 | 1 | 0% | 1,680 | 867 | -48% | 0 | 0 | — |
case-04 | pass→pass | 10,205 | 1,704 | -83% | 1 | 1 | 0% | 1,948 | 860 | -56% | 0 | 0 | — |
case-05 | fail→pass | 12,665 | 2,057 | -84% | 1 | 1 | 0% | 2,370 | 932 | -61% | 0 | 0 | — |
case-06 | pass→pass | 11,008 | 2,141 | -81% | 1 | 1 | 0% | 1,885 | 901 | -52% | 0 | 0 | — |
case-07 | fail→pass | 13,908 | 8,464 | -39% | 1 | 1 | 0% | 2,389 | 938 | -61% | 0 | 0 | — |
case-08 | pass→pass | 7,791 | 1,943 | -75% | 1 | 1 | 0% | 1,368 | 859 | -37% | 0 | 0 | — |
case-09 | fail→pass | 18,009 | 1,302 | -93% | 1 | 1 | 0% | 2,872 | 711 | -75% | 0 | 0 | — |
case-10 | fail→pass | 11,743 | 8,892 | -24% | 1 | 1 | 0% | 2,174 | 831 | -62% | 0 | 0 | — |
case-11 | pass→pass | 13,933 | 9,429 | -32% | 1 | 1 | 0% | 2,686 | 2,326 | -13% | 0 | 0 | — |
case-12 | pass→pass | 6,993 | 2,283 | -67% | 1 | 1 | 0% | 1,118 | 766 | -31% | 0 | 0 | — |
case-13 | pass→pass | 7,908 | 1,359 | -83% | 1 | 1 | 0% | 1,372 | 761 | -45% | 0 | 0 | — |
case-14 | fail→pass | 13,029 | 3,545 | -73% | 1 | 1 | 0% | 2,212 | 1,183 | -47% | 0 | 0 | — |
case-15 | fail→pass | 12,168 | 3,413 | -72% | 1 | 1 | 0% | 2,194 | 935 | -57% | 0 | 0 | — |
case-16 | fail→pass | 8,337 | 3,238 | -61% | 1 | 1 | 0% | 1,662 | 961 | -42% | 0 | 0 | — |
case-17 | pass→pass | 9,260 | 1,810 | -80% | 1 | 1 | 0% | 1,618 | 855 | -47% | 0 | 0 | — |
case-18 | pass→pass | 13,180 | 8,631 | -35% | 1 | 1 | 0% | 2,204 | 2,235 | +1% | 0 | 0 | — |
case-19 | fail→pass | 8,777 | 2,641 | -70% | 1 | 1 | 0% | 1,419 | 1,008 | -29% | 0 | 0 | — |
case-20 | pass→pass | 1,904 | 1,529 | -20% | 1 | 1 | 0% | 266 | 778 | +192% | 0 | 0 | — |
case-21 | pass→pass | 3,751 | 1,157 | -69% | 1 | 1 | 0% | 604 | 689 | +14% | 0 | 0 | — |
case-22 | pass→pass | 2,315 | 2,803 | +21% | 1 | 1 | 0% | 397 | 1,051 | +165% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 20 counted toward the lift figure. The other 2 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +36 percentage points is the difference between those two pass rates over the 20 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.