mirror of
https://github.com/garrytan/gbrain.git
synced 2026-07-30 11:22:34 +00:00
Garry Tan
08383b2943
Five red-team criticals, all closed and test-pinned: (1) two-PR gate poisoning: a same-hash committed-baseline edit must now be receipts-backed (byte-match THIS run) or the gate is inconclusive — and a receipts-backed same-hash regression still needs a justification; (2) corpus hollowing: bless-mode gold_total shrink (deleting failing fixtures / flipping holdout) now requires a justification like any adverse move; (3) continuity cross-source detection was structurally blind to writer-seeded slugs (the zero-gate was vacuous) — writer+reader slug→source maps now merge; (4) --update-baseline writes BEFORE compare (the one-shot bless flow no longer trains users to ignore exit 2) and REFUSES to write when seed failures invalidate the run; (5) 15 continuity writers' write-back gold was computed and dropped — now accumulated, so the write-back cell is flag-independent (gold 46 → 61). Plus: baselines bind their run config (holdout/llm/harness/suite — fixtures_hash covers files only), the CI gate script refuses a working-tree baseline deletion, and renderInlineTime no longer straddles UTC midnight for non-Z offsets.
135 lines
3.1 KiB
JSON
135 lines
3.1 KiB
JSON
{
|
|
"schema_version": 1,
|
|
"fixtures_hash": "76f201590dd3ad7a929e2e12efc9bf1406627b10ef4edbcfe7caf379aafd4090",
|
|
"config": {
|
|
"include_holdout": false,
|
|
"llm": false,
|
|
"harnesses": [
|
|
"claude-code",
|
|
"codex",
|
|
"openclaw"
|
|
],
|
|
"suites": [
|
|
"continuity",
|
|
"know-to-ask",
|
|
"push",
|
|
"write-back"
|
|
]
|
|
},
|
|
"cells": {
|
|
"claude-code/continuity": {
|
|
"avg_injected_tokens": 75.3333,
|
|
"continuity_rate": 1,
|
|
"source_isolation_violations": 0
|
|
},
|
|
"claude-code/know-to-ask": {
|
|
"avg_injected_tokens": 30.9247,
|
|
"false_fire_rate": 0.0233,
|
|
"know_to_ask_failure_rate": 0.15,
|
|
"source_isolation_violations": 0
|
|
},
|
|
"claude-code/push": {
|
|
"avg_injected_tokens": 38.8077,
|
|
"push_precision": 1,
|
|
"push_recall": 0.6596,
|
|
"source_isolation_violations": 0
|
|
},
|
|
"claude-code/write-back": {
|
|
"provenance_accuracy": 1,
|
|
"write_back_fidelity": 1
|
|
},
|
|
"codex/continuity": {
|
|
"avg_injected_tokens": 150.9167,
|
|
"continuity_rate": 1,
|
|
"source_isolation_violations": 0
|
|
},
|
|
"codex/know-to-ask": {
|
|
"avg_injected_tokens": 40.7123,
|
|
"false_fire_rate": 0,
|
|
"know_to_ask_failure_rate": 0.15,
|
|
"source_isolation_violations": 0
|
|
},
|
|
"codex/push": {
|
|
"avg_injected_tokens": 48.5865,
|
|
"push_precision": 1,
|
|
"push_recall": 0.4468,
|
|
"source_isolation_violations": 0
|
|
},
|
|
"codex/write-back": {
|
|
"provenance_accuracy": 1,
|
|
"write_back_fidelity": 1
|
|
},
|
|
"openclaw/continuity": {
|
|
"avg_injected_tokens": 75.3333,
|
|
"continuity_rate": 1,
|
|
"source_isolation_violations": 0
|
|
},
|
|
"openclaw/know-to-ask": {
|
|
"avg_injected_tokens": 33.7945,
|
|
"false_fire_rate": 0,
|
|
"know_to_ask_failure_rate": 0.15,
|
|
"source_isolation_violations": 0
|
|
},
|
|
"openclaw/push": {
|
|
"avg_injected_tokens": 44.1346,
|
|
"push_precision": 1,
|
|
"push_recall": 0.8085,
|
|
"source_isolation_violations": 0
|
|
},
|
|
"openclaw/write-back": {
|
|
"provenance_accuracy": 1,
|
|
"write_back_fidelity": 1
|
|
}
|
|
},
|
|
"counts": {
|
|
"claude-code/continuity": {
|
|
"gold_total": 12,
|
|
"gold_failed": 0
|
|
},
|
|
"claude-code/know-to-ask": {
|
|
"gold_total": 146,
|
|
"gold_failed": 11
|
|
},
|
|
"claude-code/push": {
|
|
"gold_total": 94,
|
|
"gold_failed": 32
|
|
},
|
|
"claude-code/write-back": {
|
|
"gold_total": 58,
|
|
"gold_failed": 0
|
|
},
|
|
"codex/continuity": {
|
|
"gold_total": 12,
|
|
"gold_failed": 0
|
|
},
|
|
"codex/know-to-ask": {
|
|
"gold_total": 146,
|
|
"gold_failed": 9
|
|
},
|
|
"codex/push": {
|
|
"gold_total": 94,
|
|
"gold_failed": 52
|
|
},
|
|
"codex/write-back": {
|
|
"gold_total": 58,
|
|
"gold_failed": 0
|
|
},
|
|
"openclaw/continuity": {
|
|
"gold_total": 12,
|
|
"gold_failed": 0
|
|
},
|
|
"openclaw/know-to-ask": {
|
|
"gold_total": 146,
|
|
"gold_failed": 9
|
|
},
|
|
"openclaw/push": {
|
|
"gold_total": 94,
|
|
"gold_failed": 18
|
|
},
|
|
"openclaw/write-back": {
|
|
"gold_total": 58,
|
|
"gold_failed": 0
|
|
}
|
|
}
|
|
}
|