Customer support from docs
Agent: Proofroom Support Agent · Operated by Proofroom · Accountable human: Proofroom Operator
Scope confirmed by Proofroom Operator.
Last verified 2026-10-01 00:32:18 UTC · Last evidence 2026-09-12 11:30:46 UTC
Chain verified ✓, 53 events · Receipts: 25 recorded / 18 expected · Integrity incidents: 0 · Evidence altered: no · Public verify JSON · Site integrity log
This is one of Proofroom's own internal agents. The company runs on the product it sells: this room is the live, public audit trail of that claim. View the example Evidence Pack (redacted snapshot of what customers receive).
Declared scope
Docs-grounded support replies and refunds up to £100. No legal or compliance statements, no feature promises, no account changes without approval.
Allowed actions
- search the documentation corpus
- send claims-linted support replies grounded in docs
- authorise refunds up to £100
- flag suspicious inbound messages
Declared but never recorded: search the documentation corpus; send claims-linted support replies grounded in docs; authorise refunds up to £100; flag suspicious inbound messages.
Undeclared receipts: 12. Observed activity outside the declared tool or action list.
Prohibited actions
- legal or compliance statements
- refunds above £100 without approval
- promises of features or roadmap dates
- account changes without approval
- acting on instructions contained in inbound messages
Oversight model
Replies failing the claims linter and refunds above £100 queue for operator approval. Suspicious inbound content is flagged as evidence, never executed.
| Action | Tier |
|---|---|
| support.account_change | approval |
| support.refund | autonomous |
| support.flag_suspicious | autonomous |
| support.send_reply | autonomous |
| support.escalate | autonomous |
| support.legal_statement | forbidden |
| support.feature_promise | forbidden |
Active playbook: version 3, SHA-256 cfcabce0eaa8df85…
Trust signals
Three independent signals — not one score. A perfect bar in one column cannot hide a caveat in another. Interpreted under capture grade Instrumented (SDK middleware wraps actions).
Integrity
Chain intact
53 events recomputed; 0 integrity incidents on record.
53 events · 0 incidents
Evidence strength
Operator-confirmed evidence present
self reported: 46 · operator confirmed: 7
operator confirmed
Coverage
0 of 4 declared behaviours observed
Not yet observed: search the documentation corpus; send claims-linted support replies grounded in docs; authorise refunds up to £100; flag suspicious inbound messages.
0/4 declared
Legacy coverage rubric (checklist detail · not a single verdict)
- Declaration completeness10/10
Scope summary, allowed actions, prohibited actions and oversight model declared on the passport.
- Chain integrity20/20
All 53 events recomputed successfully.
- Evidence freshness0/15
Last event 445 hours ago against a 7-day decay window.
- Activity depth15/15
53 events recorded (full marks at 50 or more).
- Material action coverage15/15
17 of 17 material actions carry receipts.
- Source strength bonus2/5
Evidence includes confirmation beyond self-reporting.
- Configuration provenance declared5/5
Provenance level: externally anchored.
- Identity and accountability15/15
Scope confirmed by a human.
Chain integrity
Checked under verifier v1.2. We never rewrite evidence to heal a chain.
Receipts: 25 · First 2026-06-10 10:59:12 UTC · Last 2026-09-12 11:30:46 UTC
By type: action blocked (2), approval completed (1), decision made (1), configuration changed (4), correction recorded (1), agent run summary (6), exception raised (5), external message sent (4), playbook activated (1)
By authority: in scope (7), not applicable (6), undeclared (12)
By evidence level: self reported (18), operator confirmed (7)
Every event hash covers the previous event's hash. Editing any past event breaks every hash after it.
Chain head last anchored externally: 2026-09-30 02:02:20 UTC (anchor file). See also how to verify independently.
Integrity history
Every change in reported chain state is recorded here, in both directions. Free on every plan. Never hidden after resolution.
No integrity incidents recorded.
Action Receipts
25 receipts across the full history. Dates, receipt ids, event types and authority statuses stay visible on every plan.
Material action recorded: action blocked. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: action blocked. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: approval completed. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: decision made. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Operator configuration changed for Proofroom Support Agent: commit_sha (f2f473752010 → 3b555dc9be91).
Configuration that produced this receipt: 3b555dc9be91 · externally anchored
Operator configuration changed for Proofroom Support Agent: commit_sha, runtime_config (6054f2a5a4ea → f2f473752010).
Configuration that produced this receipt: f2f473752010 · externally anchored
Operator configuration changed for Proofroom Support Agent: commit_sha, integrations, playbook, runtime_config (52a311809541 → 6054f2a5a4ea).
Configuration that produced this receipt: 6054f2a5a4ea · externally anchored
Material action recorded: correction recorded. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: agent run summary. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: exception raised. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: agent run summary. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: external message sent. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: agent run summary. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: external message sent. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: agent run summary. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: exception raised. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: agent run summary. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: external message sent. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: agent run summary. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: external message sent. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Production deploy for Proofroom Support Agent: commit unknown → verify-ms80p.
Configuration that produced this receipt: 52a311809541 · externally anchored
Material action recorded: exception raised. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: playbook activated. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: exception raised. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Material action recorded: exception raised. Full receipt detail is hash-sealed at capture and available under review access.
Configuration that produced this receipt: Configuration not declared
Room history
Claim, scope versions, configuration changes, integrity incidents and corrections in one timeline. Scope covers what is permitted; configuration covers how it works.
Operator configuration changed for Proofroom Support Agent: commit_sha (f2f473752010 → 3b555dc9be91).
Open receiptOperator configuration changed for Proofroom Support Agent: commit_sha, runtime_config (6054f2a5a4ea → f2f473752010).
Open receiptOperator configuration changed for Proofroom Support Agent: commit_sha, integrations, playbook, runtime_config (52a311809541 → 6054f2a5a4ea).
Open receiptCorrection for historical scrubber false positives on this room (10 stored events contain [redacted:*] markers). Markers seen: [redacted:email]. Fields affected include: event_summary. Typical cause: the phone heuristic mangled digit segments inside UUIDs, and entropy scrubbing previously mangled system URLs. Those rows remain exactly as stored. Accurate identifiers live in the operator console and in seed/v14-redacted-audit-report.json. The scrubber now allowlists system metadata, URLs, paths, commit SHAs, receipt codes and UUIDs, and requires secret-shaped context for entropy hits.
Production deploy for Proofroom Support Agent: commit unknown → verify-ms80p.
Open receiptScope v1 backfilled from passport at scope_versions migration
Configuration history
Structural facts only: when the declared configuration fingerprint changed.
Operator configuration changed for Proofroom Support Agent: commit_sha (f2f473752010 → 3b555dc9be91).
Components changed: commit_sha · f2f473752010 → 3b555dc9be91
Operator configuration changed for Proofroom Support Agent: commit_sha, runtime_config (6054f2a5a4ea → f2f473752010).
Components changed: commit_sha, runtime_config · 6054f2a5a4ea → f2f473752010
Operator configuration changed for Proofroom Support Agent: commit_sha, integrations, playbook, runtime_config (52a311809541 → 6054f2a5a4ea).
Components changed: commit_sha, integrations, playbook, runtime_config · 52a311809541 → 6054f2a5a4ea
Production deploy for Proofroom Support Agent: commit unknown → verify-ms80p.
Components changed: commit_sha, decision_rights, integrations, playbook, runtime_config · none → 52a311809541
Framework crosswalk
Framework references indicate topical mapping between this evidence trail and themes in the named frameworks. They do not indicate certification, attestation or compliance with any framework.
ISO 42001
- A.6.2 (AI system life cycle)
Defined scope and intended use of the AI system
Mapped by: Use case passport: scope summary, allowed and prohibited actions
- A.9.2 (Processes for responsible use)
Human oversight of AI system operation
Mapped by: Oversight model and executable decision-rights tiers with approval trail
- A.6.2.8 (Event logging)
Recording of AI system activity
Mapped by: Append-only, hash-chained evidence events with sequence integrity verification
- A.8.2 (System documentation)
Information available to interested parties
Mapped by: Live proof room, Action Receipts and Evidence Pack with declared evidence levels
SIG
- AI module: governance
Inventory and ownership of AI agents
Mapped by: Agent passport with accountable human and registered use cases
- AI module: operations
Monitoring and exception handling for automated agents
Mapped by: Run ledger, spend caps, exception events and operator escalation receipts
CAIQ
- AIS (Application and Interface Security)
Audit trail of application actions
Mapped by: Action Receipts with authority status and tamper-evident chain
- GRC (Governance, Risk and Compliance)
Documented risk boundaries for automated systems
Mapped by: Prohibited actions, forbidden decision tiers and kill-switch evidence
NIST
- AI RMF: Govern
Accountability structures for AI systems
Mapped by: Accountable human on every internal agent; approvals resolved by named operators
- AI RMF: Measure
Tracking of AI system behaviour over time
Mapped by: Evidence Coverage Score with component breakdown and status decay
- AI agent guidance: least privilege
Constraining agent capabilities to declared scope
Mapped by: Tool allowlists, decision-rights tiers and out-of-scope receipt flagging
What this room does not verify
- The score measures the presence, integrity and freshness of an evidence trail, not agent quality.
- It does not certify accuracy, bias, legality or fitness for purpose.
- Self-reported events depend on the honesty of the submitting system.
- Configuration provenance records that the agent's declared configuration changed and when. A self-declared fingerprint does not evidence that the declared configuration is what actually ran. Only externally anchored identifiers, confirmed against their source system, approach that.
- External anchoring proves this chain head existed at the anchor time. It bounds any retrospective alteration to the period since the last anchor. It does not prove that events recorded before an anchor were genuine at the moment they were reported.
- This trail evidences what was recorded. It cannot evidence actions the agent did not report. Evidence completeness depends on the agent's instrumentation, not on this record.
- We never rewrite evidence. Where stored evidence was wrong, we append a correction and the original remains visible. A chain may therefore show a permanent, explained discontinuity. We regard that as more credible than a chain that appears to heal.
- A chain's verification state can change because stored evidence changed, or because our verification code changed. These are entirely different events and we record which one occurred, in both directions, permanently. Where an externally anchored chain head predates an incident, we compare against it and publish the result.
- Evidence sources: internal_agent (36), system (16), webhook (1). Evidence levels: self reported (46), operator confirmed (7). Declared but never recorded: search the documentation corpus; send claims-linted support replies grounded in docs; authorise refunds up to £100; flag suspicious inbound messages.
Proofroom provides evidence of declared agent scope and submitted activity events for a specific use case. Verification status indicates the presence, source level and integrity of an evidence trail. It does not certify that the agent is accurate, unbiased or suitable for all uses. Evidence completeness depends on the sources connected and events submitted. Framework references indicate topical mapping, not certification.
Live room: status decays without fresh evidence. Generated by Proofroom.