Auditors don't argue with hashes. The locker is an append-only, sha-256-chained store of every screenshot, request/response pair, command log, and exploit artifact the agent produced — bound to the run, the phase, and the control it satisfies, and signed the moment the engagement ends.
A traditional pen test deliverable is a document written after the fact. The locker is the run itself: every command the agent issued, every response it observed, every screenshot it captured, stored as discrete artifacts with their own hash, timestamp, phase, and control reference.
Nothing is edited into it later. Artifacts are appended as the agent works, each one chained to the previous, and the whole chain is signed when the engagement seals. Change one byte of one artifact and verification fails — loudly, and on the auditor's machine, not ours.
Four mechanics, applied to every artifact the agent produces, with no operator in the loop.
As each phase runs, the artifact is captured at the moment it exists: the raw request and response, the screenshot, the command and its output, the manifest of what was enumerated.
A sha-256 digest is computed over the artifact bytes plus its metadata — run id, phase, control reference, timestamp. The digest is what the chain records; the bytes stay immutable beside it.
Each entry includes the digest of the entry before it, so the locker is append-only by construction. Removing, reordering, or rewriting an artifact breaks every link that follows it.
When the engagement ends — completed or aborted — the head of the chain is signed with an Ed25519 key. The signature covers the whole history, so no one, Pentr included, can rewrite a sealed locker.
Everything the agent touches becomes an entry. Nothing is summarised away.
Full-page and element captures at the moment of exploitation, including the rendered response of a successful chain, not a reconstruction of it.
Raw HTTP pairs — headers, body, status, timing — for every probe that produced a finding, so a reviewer can read the proof line by line.
The full transcript of what the agent ran and what came back, including the exploit chain scripts, with stdout and stderr preserved.
Hosts resolved, services fingerprinted, routes discovered, admin paths flagged — the machine-readable output of each reconnaissance phase.
A digest of the in- and out-of-scope lists and the rules of engagement, carried in the signed manifest. Anyone told the scope can recompute it; change the scope afterwards and it no longer matches.
Severity, CVSS, CVE where one applies, the affected target, and pointers to every artifact that proves it. The finding is a view over the evidence, never a substitute.
Verification is a local operation against an exported locker. No Pentr account, no API call, no network. If the verifier says the chain is intact, that statement doesn't depend on trusting us — it depends on the key.
The mapping is applied when the artifact is captured, not stitched together the week before the audit.
Send an auditor a scoped, read-only link to a single sealed locker. They browse artifacts and download the export without a Pentr account or a seat.
30 days on Entry, one year of signed retention on Team, and configurable retention on Enterprise — including customer-managed storage for regulated estates.
Take the whole locker with you: artifacts, chain, and seal in one archive, plus findings as CSV, PDF, or SARIF for your issue tracker.
Run an engagement against a scope you control and walk away with a locker your auditor can verify on their own machine.
Pentr