Observed build evidence · public-safe projection
Inspect what the binlog boundary retained—not the binlog itself.
This proof packet records a tiny synthetic product smoke: explicit input, exact commit binding, allowlisted evidence, deterministic projection, and attached non-claims.
Public claim level: demo. The raw artifact, digest, messages, properties, environment values, command lines, credentials, private identities, and machine-local paths are not published.
Evidence identity
One supplied artifact, one bounded rule family.
Observation rule
build.msbuild-binlog.observation.v1 · Tier2Structural · observed-boundedGap rule
build.msbuild-binlog.gap.v1 · Tier4Unknown · observed-partialExtractor
MsBuildBinlogExtractor · msbuild-binlog/0.1.0Synthetic dogfood result
The useful result is intentionally small.
Two scans over the same explicit artifact and declared commit produced the same categorical evidence.
artifact observation1 · recorded result succeeded · synthetic artifact spanproject observation1 · repository-relative identity onlyproject-reference observations0 · no absence conclusiondiagnostic observations0 · no defect-absence conclusiongaps0 for this bounded sample · not a completeness claimSafety checks
The packet validates the projection boundary.
Repeated scans
2, with deterministic projected facts.Commit bindingEvery projected fact retained the declared repository commit.
Artifact spanEvery projected fact used a synthetic artifact identity.
Protected markers
0 matches across the five standard outputs.Raw artifactDeleted after local validation and not part of this public packet.
Public claim boundary
A recorded build result is not a release conclusion.
- The packet does not authenticate or attest the binary log, or prove it was produced from the declared commit.
- A recorded
succeededresult does not prove tests passed, the repository was clean, deployment occurred, runtime behavior was correct, or a release is approved or safe. - Project and graph evidence does not prove runtime reachability. Missing diagnostics do not prove that defects are absent.
- The result is one synthetic checked-in sample, not a completeness, compatibility, or performance benchmark.
- TraceMap core scanning and reduction use no LLM calls, MCP analysis, embeddings, vector databases, or prompt-based classification.
Follow the evidence