Download a single-family reference pack for NIST 800-171 §3.6 (Incident Response), run a readable stdlib-only Python verifier, and independently check whether this copy still matches the manifest and evidence chain recorded at creation.
The proof and the verifier output are the design. Start with the files, not a claim.
verify/verify_standalone.py
tar -xzf proof_IR_2026-05-21.tar.gz cd proof_IR_2026-05-21 python3 verify/verify_standalone.py .
[*] Verifying: . [*] Verifying manifest checksums... OK (10/10 files) [*] Verifying hash chain... OK (6 links) [*] No integrity tag present on bundle. [+] VERIFICATION PASSED -- bundle integrity confirmed.
echo "x" >> evidence/INDEX.csv python3 verify/verify_standalone.py .
[*] Verifying manifest checksums... FAILED
- Hash mismatch: evidence/INDEX.csv (expected db98c66e8475c8f9..., got 90b63516bdb204ce...)
[*] Verifying hash chain... OK (6 links)
[*] No integrity tag present on bundle.
[!] VERIFICATION FAILED -- evidence may be tampered.
The verifier names the offending file with expected vs actual
SHA-256 hashes. Anything other than exit code 0 on
the clean pack means this copy no longer matches the checks
recorded when the packet was created.
Note: evidence/INDEX.csv is covered by the manifest
layer but is not itself a chained artifact, so this demo shows the
manifest catch. Modifying a chained artifact under
evidence/ breaks the Layer-2 chain instead.
Two offline checks, both inspectable from the package itself.
SHA-256 hash chain over the IR evidence artifacts. Each link binds to the previous; a naive edit without regenerating the packet leaves the recorded chain out of sync and the verifier reports the broken link.
Observed count: 6 links
SHA-256 of every content file in the pack. Append one byte to any file without regenerating the packet and the manifest layer reports that specific file with a hash mismatch on the next verify run.
Observed count: 10 / 10 files
One Incident Response family. Three controls. Evidence and verification files in one portable package.
assessment/SSP.json
The 3 IR SSP narratives (3.6.1 / 3.6.2 / 3.6.3), every citation resolves to a real file present in this pack — no placeholder text.
assessment/POAM.csv
IR-family Plan of Action and Milestones rows.
assessment/assessment-results.json
IR-family OSCAL 1.1.2 findings and observations.
evidence/INDEX.csv
IR evidence index with 6 artifacts and real SHA-256 content hashes.
evidence/evidence-bundle.json
Re-chained Layer-2 evidence bundle.
manifest.sha256
Package-level SHA-256 manifest with detached HMAC-SHA256 signature file.
verify/HOW_TO_VERIFY.md
Offline verification and changed-packet walkthrough.
verify/verify_standalone.py
Single-file Python verifier. This is the complete local verification surface for consistency checks.
README.md
Overview and cross-checked counts, with summary-report.txt beside it.
The same packet you just verified is the unit Hardseal Edge emits from your live workload.
The same packet you just verified — cryptographically chained and offline-verifiable with the single-file Python verifier above — is the unit Hardseal Edge emits from your live workload.
3.12.2.3.12.1 / 3.12.3.AU.L2-3.3.1 / -3.3.2 / -3.3.8.What the verifier shows: this copy matches the included manifest and evidence chain. It does not, by itself, provide origin authentication or defeat a deliberate re-package. Your assessor still decides whether your evidence is sufficient.
Next step: a 15-minute scoping call. No cost, no commitment, no outcome promise.
This is the trust spine. The boundary is part of the proof.
artifact_hash in evidence-bundle.json matches the SHA-256 of the corresponding file shipped in this pack - not a stub, not a placeholder.About the HMAC layer. The pack ships without the HMAC secret used to sign the evidence bundle at build time. The secret stays with the build operator. Third-party verification covers the included manifest and Layer-2 chain — it detects file-level changes against the manifest and chain shipped in the pack and proves internal consistency. Because the signing secret is withheld, it does not by itself provide origin authentication or defeat a sophisticated re-package (alter a file, then regenerate the manifest and chain). Public-key signature attribution (Ed25519) is on the Hardseal Core roadmap.
Scope restraint keeps the public proof honest.
In internal reference builds, the Hardseal pipeline generates structured draft packs mapped across the 110 CMMC Level 2 controls; evidence completeness varies by family depending on what the local environment exposes to the collectors. Across the full Acme reference build, 23 of the top-40 SSP narratives carry all-real citations and 17 carry explicit scope-gap placeholders, clearly marked "evidence to be collected during delivery".
For this public proof, we ship the one family (IR, Incident Response, three controls) where every cited artifact exists inside the synthetic pack and verifies cleanly, with no narrative dressed up as something it is not. The honesty doctrine is non-negotiable: a proof that needs caveats stops being a proof.
Hardseal can run one real control family through this same offline-verifiable evidence-pack process for feedback. No cost, no commitment, no outcome promise. Just pre-assessment readiness evidence you can inspect.