Defendable Offense / Pain / Fixer Corpus v1
Source-grounded offense-signal / pain-receipt corpus + defense-fixer library + SwarmCurator eval and SwarmJelly failure packs. 678 records across 6 research lanes, trust-tiered to primary enforcement sources. Salvaged from the archived defendableos-swarm-research corpus.
Asset Snapshot
- Domain
- Defense Intelligence
- Category
- Agent Instructions
- Version
- 1.0.0
- Records
- 678
- Size
- 1.4 MB
- Quality
- 90/100
- Fine-tune readiness
- 100/100
- Readiness label
- Fine-tune ready
- Validation
- passed
Intended Use
Train and evaluate SwarmCurator (grading), test SwarmJelly (failure/repair), and seed the Defendable Ledger with source-grounded offense/pain/fixer intelligence.
Not Intended Use
Not operational attack instructions (fixers are safe-high-level with evidence + human-review gates). Not legal, financial, or security advice.
Schema Preview
{
"record_id": "REC-FIN-0001",
"research_lane": "consumer_financial",
"fields": [
"classification",
"corpus_version",
"created_at",
"curator_training",
"privacy_and_safety",
"record_id",
"research_lane",
"severity",
"source",
"trust"
],
"note": "Full records in splits/offense_pain_seed_corpus.jsonl; 3 in samples/sample.jsonl."
}Example Records
[
{
"record_id": "REC-FIN-0001",
"research_lane": "consumer_financial",
"fields": [
"classification",
"corpus_version",
"created_at",
"curator_training",
"privacy_and_safety",
"record_id",
"research_lane",
"severity",
"source",
"trust"
],
"note": "Full records in splits/offense_pain_seed_corpus.jsonl; 3 in samples/sample.jsonl."
}
]Formats and Tasks
Fine-tuning Notes
Salvaged from the archived SudoSuOps/defendableos-swarm-research corpus v0.1 (deliverables 05/06/07/08) during the 2026-05-31 house consolidation. 678 records across seed corpus (255), fixer library (115), curator eval (157), jelly failure (151). Trust-tiered source-grounded; SHA256 receipts captured; full data committed in-repo.
678 records, all lines valid JSON, per-file + combined SHA256 captured and re-verified. Seed corpus 183x trust-tier-1 / 72x trust-tier-2.
Files, Hashes, and Receipts
| splits/offense_pain_seed_corpus.jsonl | jsonl | full | 616 KB | 14617a8258d9303f2a7467ff22170a9b3d3840cd285cb885cce344ae404fb94b |
| splits/defense_fixer_library.jsonl | jsonl | full | 172 KB | 6068ac4dfd3acb77b2ac6a17b91feb5529cec2ec50b6ddeaa193a82bacdd06f6 |
| splits/swarmcurator_eval_pack.jsonl | jsonl | full | 354 KB | d477d5285c267b9c84fa6d306fd518cf991f3364290f801f4a02742a91411965 |
| splits/swarmjelly_failure_pack.jsonl | jsonl | full | 293 KB | 47702273a79cd496fee36f62fa05605b7bf43d294dc789563e111e099ad61d64 |
External Storage Locations
Dataset Card Preview
--- license: defendable-community-research task_categories: classification, instruction-tuning pretty_name: Defendable Offense / Pain / Fixer Corpus v1 --- # Defendable Offense / Pain / Fixer Corpus v1 Source-grounded offense-signal / pain-receipt corpus + defense-fixer library + SwarmCurator eval and SwarmJelly failure packs. 678 records across 6 research lanes, trust-tiered to primary enforcement sources. Salvaged from the archived defendableos-swarm-research corpus. ## Datasets - Defendable Offense / Pain / Fixer Corpus v1 (defendable_offense_pain_fixer_v1) - 678 records ## Provenance This card was generated from DefendableDatasets registry metadata. Verify SHA256 hashes before training. ## Warnings - One or more datasets use a gated research license. Review before commercial use or redistribution. - License compatibility is restricted for production fine-tuning.