Update to Lukka Digital Assets Intel

{
  "description": "Process large MCP responses (>100KB) locally instead of loading into context. Covers market depth (6.8MB+), bulk address transactions (150KB+), block batch analysis (675KB+), and AML standard/enhanced reports (300-500KB+). Deploy Python processors to /tmp, pipe raw JSON through them, read concise output (~2-3KB) back into context.",
  "included_files": [
    {
      "relative_path": "references/lukka_notation.py",
      "size_in_bytes": 5592
    },
    {
      "relative_path": "references/process_aml_report.py",
      "size_in_bytes": 12614
    },
    {
      "relative_path": "references/process_block.py",
      "size_in_bytes": 8532
    },
    {
      "relative_path": "references/process_market_depth.py",
      "size_in_bytes": 5610
    },
    {
      "relative_path": "references/process_transactions.py",
      "size_in_bytes": 8939
    }
  ],
  "name": "large-data-processing",
  "skill_md_contents": "---\nname: large-data-processing\ndescription: \"Process large MCP responses (>100KB) locally instead of loading into context. Covers market depth (6.8MB+), bulk address transactions (150KB+), block batch analysis (675KB+), and AML standard/enhanced reports (300-500KB+). Deploy Python processors to /tmp, pipe raw JSON through them, read concise output (~2-3KB) back into context.\"\n---\n\n## OpenAI connection notes\n\nTool identifiers below are raw operation names. The host may add a namespace;\nuse the matching advertised Global MCP tool. Resolve `<resolved-package-root>`\nto this plugin's installed directory before using paths. Workflow names identify\npackaged skills; they are not literal Claude slash commands.\n\n# Large Data Processing\n\n**STANDALONE**: This skill provides Python processor scripts (stdlib only) that summarize large JSON files locally -- market depth, transactions, blocks, AML reports. Works on any JSON data matching the expected schemas. **SUPERCHARGED**: When Lukka MCP servers are connected, these processors prevent context overflow from oversized API responses (6.8MB market depth, 500KB AML reports).\n\n**Problem**: Some MCP responses are too large for context (6.8MB market depth, 1000 decoded txs).\n**Solution**: Save raw JSON to /tmp, process locally with Python scripts, read summary.\n\n## When to Use\n\n| Endpoint | Trigger | Size |\n|---|---|---|\n| `get_market_depth` without pair filter | Full exchange dump | 6.8MB, 1500+ pairs |\n| `get_order_book_snapshot` (single pair) | L2 book snapshot | ~3MB EVEN WITH pair filter - always pipe |\n| `get_derivative_source_details` | Derivative pair catalog | ~1.5MB (3800+ pairs) - grep for the target code, never load raw |\n| `get_historical_derivative_trades` | Derivative trade prints | 300KB+ per 1h; `limit` is per-time-page NOT per-record, so it does NOT cap size. Use smallest window + pipe |\n| `getImpliedRatesOtcFx` | OTC FX term structure | 400KB+; `limit` caps time-pages not rows (all tenors x all hours return). Narrow window + pipe |\n| `get_asset_by_id` (hosted) | Full asset detail | ~54KB raw (hosted has no summary mode) - fits context but trim if batching |\n| `get_address_transactions` verbose, limit>100 | Bulk wallet analysis | 150KB+ |\n| `get_block` + batch decode | Block-level forensics | 675KB+ (135 txs) |\n| `generateAmlStandardReport` | Full AML report | 300-500KB (ALWAYS pipe) |\n| `generateAmlEnhancedReport` | Deepest AML report | 500KB+ (ALWAYS pipe) |\n| Any MCP response | Response feels truncated or huge | >50KB |\n\n## Workflow\n\n### Step 1: Detect large response risk\n\nBefore calling, check if the query will return bulk data:\n- `get_market_depth` without `baseAssetCodes`/`counterAssetCodes` = 6.8MB\n- `get_order_book_snapshot` = ~3MB even with a single pair filter\n- `get_derivative_source_details` = ~1.5MB (grep for the pair, never load raw)\n- `get_historical_derivative_trades` / `getImpliedRatesOtcFx` = 300-400KB+; `limit` does NOT cap total size (it pages by time), so shrink the from/to window\n- `get_address_transactions` with `limit: 1000` + `verbose: \"1\"` = 150KB+\n- Batch decoding >50 transactions = 250KB+\n\n### Step 2: Deploy processor script\n\nCopy the appropriate script from this skill's references to /tmp. When invoked as a plugin skill, use `<resolved-package-root>` to resolve the plugin install path:\n\n```bash\n# Market depth\ncp \"<resolved-package-root>/skills/large-data-processing/references/process_market_depth.py\" /tmp/\n\n# Transactions\ncp \"<resolved-package-root>/skills/large-data-processing/references/process_transactions.py\" /tmp/\n\n# Block data\ncp \"<resolved-package-root>/skills/large-data-processing/references/process_block.py\" /tmp/\n\n# AML standard/enhanced report\ncp \"<resolved-package-root>/skills/large-data-processing/references/process_aml_report.py\" /tmp/\n```\n\n### Step 3: Call MCP and save raw output\n\nCall the MCP tool with `format_response: false` or `detailed: true` where available.\nSave the raw JSON response to /tmp:\n\n```bash\n# Example: save market depth response\necho '<raw_json>' > /tmp/depth.json\n```\n\n### Step 4: Process and read summary\n\n```bash\n# Market depth - top 20 most liquid pairs\npython3 /tmp/process_market_depth.py /tmp/depth.json --top 20\n\n# Transactions - full statistical summary\npython3 /tmp/process_transactions.py /tmp/txs.json --stats --top-senders 10 --dust-threshold 0.0001\n\n# Block analysis\npython3 /tmp/process_block.py /tmp/block.json --stats --sample 5\n\n# Multi-block batch (JSONL)\npython3 /tmp/process_block.py /tmp/blocks.jsonl --range-stats\n\n# AML report (Standard/Enhanced) - MANDATORY when calling those two tools\npython3 /tmp/process_aml_report.py /tmp/aml_report.json\n```\n\n### Step 5: Follow up via MCP\n\nUse the summary to identify specific items for deeper MCP queries:\n- Decode flagged transactions: `UDA___Get_transaction_for_txhash({ blockchain, txhash })`\n- Get specific pair depth: `pricing___get_market_depth({ sourceIds, baseAssetCodes, counterAssetCodes })`\n- AML flagged counterparties: run `AML___generateAmlScoringReport` on each\n\n## Script Reference\n\n### process_market_depth.py\n\n```\npython3 process_market_depth.py <file.json> [options]\n  --pair SOLN-USD          Filter specific pair\n  --top N                  Top N pairs by depth\n  --min-depth 50000        Min total depth (USD)\n  --sort depth|spread|volume   Sort order (default: depth)\n  --format table|csv|json  Output format (default: table)\n```\n\nOutput: Exchange, Pair, Bid 1%, Ask 1%, Total Depth, Spread bps, Mid Price\n\n### process_transactions.py\n\n```\npython3 process_transactions.py <file.json> [options]\n  --stats                  Summary: count, volume, date range, errors\n  --address 0x...          Target address (auto-detected if omitted)\n  --top-senders N          Top N inbound addresses\n  --top-recipients N       Top N outbound addresses\n  --min-value 0.1          Min value filter\n  --from-date 2026-01-01   Date filter (from)\n  --to-date 2026-03-12     Date filter (to)\n  --large-txs N            N largest txs (for MCP decode)\n  --dust-threshold 0.0001  Dust detection threshold\n  --format table|csv|json  Output format\n```\n\nOutput: Total txs, date range, inflow/outflow, error rate, dust count, top senders/recipients, largest txs.\n\n### process_block.py\n\n```\npython3 process_block.py <file.json> [options]\n  --stats                  Block summary: tx count, gas, time\n  --sample N               Random N tx hashes for decode\n  --range-stats            Multi-block stats (JSONL input)\n  --format table|csv|json  Output format\n```\n\nOutput: Block number, timestamp, tx count, size, gas usage, base fee, sample hashes.\n\n### process_aml_report.py\n\n```\npython3 process_aml_report.py <file.json> [options]\n  --full-risks            Include all risk factors (not just top 10)\n  --format table|json     Output format (default: table)\n```\n\nOutput (~2-3 KB): report info (address, score, grade), top risk factors, flagged counterparties, exposure summary, monitoring flags (sanctions/darknet/ransomware/mixer counts), tx stats, recommendations.\n\n**Always use this when calling** `AML___generateAmlStandardReport` or `AML___generateAmlEnhancedReport` - raw responses exceed context.\n\n## Anti-patterns\n\n- Do NOT load 6.8MB raw JSON into context - use this skill\n- Do NOT decode all 135 block transactions one by one - sample first\n- Do NOT call `get_market_depth` without pair filters unless you WANT full exchange data\n- Do NOT call `generateAmlStandardReport` or `generateAmlEnhancedReport` without the processor - payload exceeds context\n- Do NOT skip Step 2 (deploy) - scripts must exist in /tmp before use\n"
}