measurements: FIRST cluster demonstration (xiaomusic prefix-bypass path traversal) [skip ci]

search-with-proof produced its first demonstrated finding on a real CVE fix-pair end to
end on kube-ax: vuln observes, fixed does not, 0 false proofs, $0.37, 55 calls. Enabled
by the full session fix chain (parity, verify-time-only, fair-share budget, served-root
guidance, oracle prefix-bypass coverage, control-signal feedback). Payload-free record.

Co-Authored-By: Claude Opus 4.8 <[email protected]>
This commit is contained in:
Marius Ciepluch
2026-10-07 12:41:01 +02:00
co-authored by Claude Opus 4.8
parent 1ec4ee4c55
commit ecd1372568
@@ -0,0 +1,22 @@
{
"record": "2026-10-07 FIRST CLUSTER DEMONSTRATION for search-with-proof",
"pair": {"index": 13, "repo": "hanxi/xiaomusic", "family": "path", "class": "prefix-check-bypass (startswith without separator)", "language": "python"},
"image": "ghcr.io/norandom/ousast-search-task@sha256:e2c8b460...",
"side": "vulnerable",
"outcome": "demonstrated",
"verify_tasks": 18,
"model_calls": 55,
"settled_usd": 0.3686,
"wall_seconds": 2170,
"false_proofs_fixed_side": 0,
"demo_wires_served_root": true,
"finding": "The agent autonomously produced a demonstrated finding on a real CVE fix-pair end to end on the kube-ax cluster: it found the prefix-bypass sink, wired the app's served directory to ${served_root} (prompted by the new served-root control-signal hint after an earlier mis-wire), targeted the oracle's prefix-sibling canary, and the differential verifier confirmed the owned effect on the vulnerable revision and its absence on the fixed revision. Cost $0.37, 55 model calls.",
"fix_chain_that_enabled_it": [
"parity: agent targets the observable",
"verify-time-only: no explore reachability trap",
"fair-share budget: authors+verifies within budget",
"served-root guidance: wire the served dir to ${served_root}",
"oracle prefix-bypass coverage: sibling canary for startswith-without-separator vulns",
"control-signal feedback: served-root hint so the agent self-corrects wiring"
]
}