Hardproof quality report (release QA)
This page is a public summary of the first recorded Hardproof release QA run against non-fixture targets. It is intended to be reproducible: every row links to the raw scan outputs and the exact manifest used.
Evidence bundle
- How to rerun: https://github.com/x07lang/hardproof/blob/04f5d7f4a63b606e1d8ac24cc39a20365a441e45/release_qa/external_smoke_matrix/README.md · https://github.com/x07lang/hardproof/blob/04f5d7f4a63b606e1d8ac24cc39a20365a441e45/scripts/release_qa/run_external_smoke_matrix.sh
- Tool versions/commits: https://github.com/x07lang/hardproof/blob/04f5d7f4a63b606e1d8ac24cc39a20365a441e45/release_qa/external_smoke_matrix/0.4.0-beta.2/meta.json
- Full command transcript: https://github.com/x07lang/hardproof/blob/04f5d7f4a63b606e1d8ac24cc39a20365a441e45/release_qa/external_smoke_matrix/0.4.0-beta.2/command.log
- Corpus manifest: https://github.com/x07lang/hardproof/blob/04f5d7f4a63b606e1d8ac24cc39a20365a441e45/release_qa/external_smoke_matrix/external-smoke-001.json
Scan summary
Hardproof makes score truth explicit. Partial scans are numeric-but-not-publishable (score_truth_status=partial). overall_score is still computed as the effective score (matching partial_score); full scans publish overall_score once the weighted evidence gates (typically Trust) are satisfied.
Note: this evidence bundle was recorded with 0.4.0-beta.2, which used overall_score=null for partial scans. The table below shows the effective score (the value that now appears as overall_score in partial scans).
| target | status | score_truth_status | score_mode | overall_score | partial_score | evidence |
|---|---|---|---|---|---|---|
stdio x07lang-mcp | fail | partial | partial | 78 | 78 | scan.json |
streamable_http postgres-mcp demo (partial) | warn | partial | partial | 96 | 96 | scan.json |
streamable_http postgres-mcp demo (full) | warn | publishable | full | 93 | 93 | scan.json |
Corpus flow (public-sample pipeline)
The release QA run also records a minimal corpus run + render to prove the public-sample pipeline works end-to-end.
- Rendered corpus report: report.html
- Corpus index: index.json
- Corpus result: result.json
How to interpret this page
- Methodology (scoring, confidence, exclusions): /hardproof/methodology
- Report schema: /hardproof/report-format
- Security guide: /hardproof/security-guide
- Pipeline outline (draft): /hardproof/quality-report-pipeline