Evidence for your engineering decision
LSAS Platform recorded 4 governed requests across 1 policy version: 2 final outputs authorized unchanged, 0 final outputs sanitized before release authorization, and 2 withheld. These counts describe final outcomes; an earlier ingress transformation is recorded separately in scenario evidence. They do not establish application-wide detection accuracy.
Recorded evidence
The retained evidence for this evaluation organization.
10000000-0000-4000-8000-000000000003Share of retained requests · 0–100%
One final outcome per recorded request. Percentages use the full retained total, not the appendix; rounded values may not sum to 100%.
Final outcomes describe the checked output. Earlier policy decisions and permission to execute remain separate evidence.
Matching requests · shared scale 0–2
Matching recorded requests; controls can overlap and must not be added into a request total. Zero means not observed in this evidence, not a failed or unavailable control.
4 records inspected. Content controls from payload and synthetic-provider records. Action-authority observations appear in the evidence table below.
Complete retained decision history · 4 records included from 4 lifetime decisions. Counts and policy totals cover the full retained history, not a recent-page sample.
Snapshot generated 2026-11-07T06:58:36.404Z from 4 retained records. The complete recorded history is present.
Scroll horizontally to inspect every column. With a keyboard, focus the table and use the Left and Right arrow keys.
| Control | Recorded observation | Exact policy and example receipts |
|---|---|---|
| Authorize unchanged output The final content boundary authorized output without a transformation. | Observed 1 matching request | developer-data-guard@1.0.0Request 30000000-0000-4000-8000-000000000004 |
| Sanitize before the provider Ingress was transformed before the synthetic provider was invoked. | Not observed 0 matching requests | No matching retained receipt |
| Block before provider dispatch Ingress was blocked and the synthetic provider received zero calls. | Observed 2 matching requests | developer-data-guard@1.0.0Request 30000000-0000-4000-8000-000000000001Request 30000000-0000-4000-8000-000000000002 |
| Hold before provider dispatch Ingress was held and the synthetic provider received zero calls. | Not observed 0 matching requests | No matching retained receipt |
| Release with a warning The final content decision retained a warning while authorizing release. | Not observed 0 matching requests | No matching retained receipt |
| Sanitize output before release The output boundary transformed and revalidated the content before authorizing release. | Not observed 0 matching requests | No matching retained receipt |
| Block at the output boundary A payload or synthetic-provider response was blocked at egress. | Not observed 0 matching requests | No matching retained receipt |
| Hold at the output boundary A payload or synthetic-provider response was withheld at egress. | Not observed 0 matching requests | No matching retained receipt |
| Authorize the registered action The kernel authorized the fixed synthetic reader and recorded one invocation. | Observed 1 matching request | developer-data-guard@1.0.0Request 30000000-0000-4000-8000-000000000003 |
| Deny an unauthorized action The kernel denied the fixed action attempt and recorded zero invocations. | Not observed 0 matching requests | No matching retained receipt |
Counts are matching recorded requests, not distinct test cases, detection accuracy, or a suite pass rate. A missing observation means this retained evidence does not demonstrate that control; it does not establish that the control failed or is unavailable. Local suite assertions and application handler behavior are not persisted here.
developer-data-guard@1.0.0Deterministic JSON data-boundary evaluation for supported personal-data, credential and prompt-injection patterns. Not a compliance certification.
Recorded engine timing varies by route: payload requests measure the initial validator pass; synthetic-provider scenarios span ingress, the local adapter (including its pre-dispatch authority check), and egress up to the last policy evaluation. Final transformation checks can fall outside that interval. Action tests include content checks, local kernel execution, intermediate authority reads, and kernel audit writes, before final publication checks. Excludes network, authentication, final database commit, and work performed by your application. Mixed-route summaries are not comparable latency benchmarks. A recorded zero reflects timer precision, not zero end-to-end latency. This is not a load benchmark.
Declare expectations for representative synthetic cases, run the reusable test kit in CI, and reconcile its local results with these committed request records. Local assertions and receiver behavior remain separate evidence.
Read the Developer Evaluation guide →Use the observed controls and evidence gaps to identify which application data surfaces need a scoped Discovery Scan. The evaluation does not establish estate-wide coverage.
Discuss scope with LSAS Platform →Bring the local brief's application purpose, integration point, open questions, and these exact policy references into organization onboarding and alignment review. A tenant pilot establishes the permitted runtime boundary; this evaluation grants no approval or deployment authority.
Discuss scope with LSAS Platform →Policy names describe the exact recorded version. Organizational baseline approval, accountable approver and completed human review are not established by these fixed-policy evaluation receipts.
RELEASED means authorized for return. TRANSFORMED means changed and revalidated before release authorization. BLOCKED and HELD mean protected content was withheld. A commit is not a downstream delivery acknowledgement.
Scroll horizontally to inspect every column and the complete fingerprints. With a keyboard, focus the table and use the Left and Right arrow keys.
| What happened and why | Request / recorded at UTC | Exact policy | Recorded engine timing |
|---|---|---|---|
| BlockedRelease denied Input blocked before provider dispatch. Input rule details were not retained for this request. The provider was not invoked; no provider output was evaluated. Final decision: BLOCKED · ingress Rules in this decision: PROMPT_INJECTION_HEURISTIC Synthetic scenario: block-input (original example) Input remediation details were not retained. | 30000000-0000-4000-8000-0000000000012026-10-29T18:40:14.411Z | Developer Data Guarddeveloper-data-guard@1.0.0 | 0 ms (rounded) |
| Receipt fingerprints · 30000000-0000-4000-8000-000000000001 No released artifact fingerprint. Protected content was withheld. | |||
| BlockedRelease denied Input blocked before provider dispatch. Input rule details were not retained for this request. The provider was not invoked; no provider output was evaluated. Final decision: BLOCKED · ingress Rules in this decision: PROMPT_INJECTION_HEURISTIC Synthetic scenario: block-input (original example) Input remediation details were not retained. | 30000000-0000-4000-8000-0000000000022026-10-29T18:06:55.808Z | Developer Data Guarddeveloper-data-guard@1.0.0 | 1 ms |
| Receipt fingerprints · 30000000-0000-4000-8000-000000000002 No released artifact fingerprint. Protected content was withheld. | |||
| AllowedRegistered synthetic read authorized 1 recorded synthetic reader invocation. The final outcome governs whether its output was authorized for return. Final decision: ALLOW · egress Rules in this decision: None recorded Action case: allowed-read · AUTHORIZED Recorded grant: Execution receipt: computed-unreleased | 30000000-0000-4000-8000-0000000000032026-10-27T18:47:17.044Z | Developer Data Guarddeveloper-data-guard@1.0.0 | 2142 ms |
Receipt fingerprints · 30000000-0000-4000-8000-000000000003
| |||
| AllowedOutput authorized unchanged Input authorized for provider dispatch. Input rule details were not retained for this request. No rule matches were recorded at the final output boundary. This does not establish that all risks were absent. Final decision: ALLOW · egress Rules in this decision: None recorded Synthetic scenario: safe-release (original example) Input remediation details were not retained. | 30000000-0000-4000-8000-0000000000042026-10-27T18:46:48.707Z | Developer Data Guarddeveloper-data-guard@1.0.0 | 380 ms |
Receipt fingerprints · 30000000-0000-4000-8000-000000000004
| |||
Artifact fingerprints are SHA-256 of JSON.stringify of the authorized returned data, preserving serialization order. Grant fingerprints describe historical execution, not reusable authority; a computed receipt alone does not authorize release. Hashes are not anonymization guarantees.
Artifact fingerprints are SHA-256 of JSON.stringify of the authorized returned data, preserving serialization order. Withheld content has no released artifact fingerprint. Hashes are not anonymization guarantees.
{
"schemaVersion": 2,
"title": "LSAS Platform Developer Evaluation - evidence brief",
"generatedAt": "2026-11-07T06:58:36.404Z",
"scope": "Recorded deterministic content governance, synthetic-provider boundaries, and fixed synthetic-reader action authority. No external model provider, arbitrary tool, or production-workflow execution.",
"evaluation": {
"id": "10000000-0000-4000-8000-000000000001",
"userId": "10000000-0000-4000-8000-000000000002",
"tenantId": "10000000-0000-4000-8000-000000000003",
"appId": "10000000-0000-4000-8000-000000000004",
"status": "ACTIVE",
"createdAt": "2026-10-27T18:41:47.036Z",
"activationDeadline": "2026-11-03T18:41:47.036Z",
"activatedAt": "2026-10-27T18:46:48.707Z",
"expiresAt": "2026-11-10T18:46:48.707Z",
"decisionCount": 4,
"decisionLimit": 10000,
"rateLimit": 30,
"policyPackId": "developer-data-guard",
"policyPackVersion": "1.0.0",
"policyRevision": 1
},
"summary": {
"retainedDecisions": 4,
"lifetimeDecisions": 4,
"complete": true,
"firstDecisionAt": "2026-10-27T18:46:48.707Z",
"lastDecisionAt": "2026-10-29T18:40:14.411Z",
"outcomes": [
{
"outcome": "BLOCKED",
"count": 2
},
{
"outcome": "RELEASED",
"count": 2
}
],
"policies": [
{
"policyPackId": "developer-data-guard",
"policyPackVersion": "1.0.0",
"count": 4
}
],
"latency": {
"samples": 4,
"averageMs": 630.75,
"p50Ms": 190.5,
"p95Ms": 1877.7,
"maxMs": 2142
}
},
"complete": true,
"outcomes": [
{
"outcome": "RELEASED",
"label": "Allowed",
"title": "Output authorized unchanged",
"description": "LSAS Platform authorized the final checked output for return without transforming that output. A synthetic scenario may still have transformed the earlier request at ingress. This applies to the checks enabled by the recorded policy.",
"short": "Final output authorized without transformation",
"tone": "allowed",
"count": 2
},
{
"outcome": "TRANSFORMED",
"label": "Sanitized",
"title": "Sanitized before release",
"description": "LSAS Platform transformed the payload on the server, revalidated the result, and authorized the transformed artifact for return.",
"short": "Transformed and revalidated before authorization",
"tone": "sanitized",
"count": 0
},
{
"outcome": "BLOCKED",
"label": "Blocked",
"title": "Release denied",
"description": "The recorded content or action controls denied release. The protected output was withheld; your application should use the returned denial instead of forwarding the original content.",
"short": "Protected output withheld",
"tone": "blocked",
"count": 2
},
{
"outcome": "HELD",
"label": "Held",
"title": "Release conditions were not met",
"description": "The release boundary withheld the payload. This evaluation does not provide a review or override action that bypasses the hold.",
"short": "Release conditions were not met; content withheld",
"tone": "held",
"count": 0
}
],
"policies": [
{
"policyPackId": "developer-data-guard",
"policyPackVersion": "1.0.0",
"count": 4,
"name": "Developer Data Guard",
"description": "Deterministic JSON data-boundary evaluation for supported personal-data, credential and prompt-injection patterns. Not a compliance certification.",
"coverage": [
"Email, phone and supported identifier patterns",
"Structured credentials and known secret patterns",
"Known prompt-injection patterns"
],
"limitations": [
"Pattern-based checks can miss sensitive data.",
"No model call, destination delivery or full security assessment."
],
"enabledValidators": [
"pii-phi",
"prompt-injection",
"security-secrets"
]
}
],
"controlCoverage": {
"schemaVersion": 1,
"generatedAt": "2026-11-07T06:58:36.404Z",
"retainedDecisions": 4,
"lifetimeDecisions": 4,
"recordsInspected": 4,
"retainedSetComplete": true,
"complete": true,
"recordsWithoutBoundaryDetails": 0,
"scenarioRequests": 3,
"actionRequests": 1,
"actionInvocations": 1,
"actionDenials": 0,
"controls": [
{
"id": "output-release",
"title": "Authorize unchanged output",
"description": "The final content boundary authorized output without a transformation.",
"group": "Content",
"count": 1,
"state": "OBSERVED",
"requestIds": [
"30000000-0000-4000-8000-000000000004"
],
"policyVersions": [
"developer-data-guard@1.0.0"
],
"additionalPolicyVersions": 0
},
{
"id": "ingress-transform",
"title": "Sanitize before the provider",
"description": "Ingress was transformed before the synthetic provider was invoked.",
"group": "Content",
"count": 0,
"state": "NOT_OBSERVED",
"requestIds": [],
"policyVersions": [],
"additionalPolicyVersions": 0
},
{
"id": "ingress-block",
"title": "Block before provider dispatch",
"description": "Ingress was blocked and the synthetic provider received zero calls.",
"group": "Content",
"count": 2,
"state": "OBSERVED",
"requestIds": [
"30000000-0000-4000-8000-000000000001",
"30000000-0000-4000-8000-000000000002"
],
"policyVersions": [
"developer-data-guard@1.0.0"
],
"additionalPolicyVersions": 0
},
{
"id": "ingress-hold",
"title": "Hold before provider dispatch",
"description": "Ingress was held and the synthetic provider received zero calls.",
"group": "Content",
"count": 0,
"state": "NOT_OBSERVED",
"requestIds": [],
"policyVersions": [],
"additionalPolicyVersions": 0
},
{
"id": "output-warning",
"title": "Release with a warning",
"description": "The final content decision retained a warning while authorizing release.",
"group": "Content",
"count": 0,
"state": "NOT_OBSERVED",
"requestIds": [],
"policyVersions": [],
"additionalPolicyVersions": 0
},
{
"id": "output-transform",
"title": "Sanitize output before release",
"description": "The output boundary transformed and revalidated the content before authorizing release.",
"group": "Content",
"count": 0,
"state": "NOT_OBSERVED",
"requestIds": [],
"policyVersions": [],
"additionalPolicyVersions": 0
},
{
"id": "output-block",
"title": "Block at the output boundary",
"description": "A payload or synthetic-provider response was blocked at egress.",
"group": "Content",
"count": 0,
"state": "NOT_OBSERVED",
"requestIds": [],
"policyVersions": [],
"additionalPolicyVersions": 0
},
{
"id": "output-hold",
"title": "Hold at the output boundary",
"description": "A payload or synthetic-provider response was withheld at egress.",
"group": "Content",
"count": 0,
"state": "NOT_OBSERVED",
"requestIds": [],
"policyVersions": [],
"additionalPolicyVersions": 0
},
{
"id": "action-authorized",
"title": "Authorize the registered action",
"description": "The kernel authorized the fixed synthetic reader and recorded one invocation.",
"group": "Action",
"count": 1,
"state": "OBSERVED",
"requestIds": [
"30000000-0000-4000-8000-000000000003"
],
"policyVersions": [
"developer-data-guard@1.0.0"
],
"additionalPolicyVersions": 0
},
{
"id": "action-denied",
"title": "Deny an unauthorized action",
"description": "The kernel denied the fixed action attempt and recorded zero invocations.",
"group": "Action",
"count": 0,
"state": "NOT_OBSERVED",
"requestIds": [],
"policyVersions": [],
"additionalPolicyVersions": 0
}
],
"interpretation": "Counts are matching recorded requests, not distinct test cases, detection accuracy, or a suite pass rate. A missing observation means this retained evidence does not demonstrate that control; it does not establish that the control failed or is unavailable. Local suite assertions and application handler behavior are not persisted here."
},
"recordedRuleMatches": [
{
"id": "PROMPT_INJECTION_HEURISTIC",
"title": "Instruction override pattern",
"explanation": "An enabled check matched a phrase associated with overriding or revealing instructions. Phrase matching does not cover every prompt-injection technique.",
"category": "Prompt instructions",
"count": 2
}
],
"timingScope": "Recorded engine timing varies by route: payload requests measure the initial validator pass; synthetic-provider scenarios span ingress, the local adapter (including its pre-dispatch authority check), and egress up to the last policy evaluation. Final transformation checks can fall outside that interval. Action tests include content checks, local kernel execution, intermediate authority reads, and kernel audit writes, before final publication checks. Excludes network, authentication, final database commit, and work performed by your application. Mixed-route summaries are not comparable latency benchmarks.",
"limitations": [
"These are deterministic checks on the recorded requests, not a certification or a guarantee that every sensitive value or harmful instruction is detected.",
"Persisted evidence proves the release decision committed. It does not prove that a downstream application received or correctly used the returned content.",
"Payload requests use deterministic checks; provider scenarios and the action lab use fixed in-process synthetic adapters. No external model provider, external connector, arbitrary tool, or production workflow is exercised.",
"Observed control counts describe committed requests. They do not verify a local contract suite's assertions, establish detection accuracy, or grant application deployment or review authority.",
"Payload bodies are not retained in this report. Compare request IDs with your local terminal or the transient console scenario result to inspect actual returned content."
],
"nextSteps": [
{
"title": "Keep an application contract suite",
"description": "Declare expectations for representative synthetic cases, run the reusable test kit in CI, and reconcile its local results with these committed request records. Local assertions and receiver behavior remain separate evidence.",
"path": "/console/evaluation?tenantId=10000000-0000-4000-8000-000000000003#application-suite"
},
{
"title": "Connect the evidence to Discovery",
"description": "Use the observed controls and evidence gaps to identify which application data surfaces need a scoped Discovery Scan. The evaluation does not establish estate-wide coverage.",
"path": "/contact"
},
{
"title": "Prepare your application for a pilot",
"description": "Bring the local brief's application purpose, integration point, open questions, and these exact policy references into organization onboarding and alignment review. A tenant pilot establishes the permitted runtime boundary; this evaluation grants no approval or deployment authority.",
"path": "/contact"
}
],
"decisions": [
{
"id": "20000000-0000-4000-8000-000000000001",
"requestId": "30000000-0000-4000-8000-000000000001",
"decisionEventId": "40000000-0000-4000-8000-000000000001",
"createdAt": "2026-10-29T18:40:14.411Z",
"decision": "BLOCKED",
"overallScore": 18,
"tier": "HIGH",
"domainScores": [
{
"score": 18,
"domain": "SECURITY",
"maxScore": 18
}
],
"ruleIds": [
"PROMPT_INJECTION_HEURISTIC"
],
"validatorIds": [
"pii-phi",
"prompt-injection",
"security-secrets"
],
"remediationTypes": [
"BLOCK"
],
"timings": {
"totalMs": 0,
"upstreamMs": 0,
"preflightMs": 0,
"postflightMs": 0
},
"policyPackId": "developer-data-guard",
"policyPackVersion": "1.0.0",
"model": "lsas-evaluation-fixture-1",
"provider": "lsas-evaluation-recording",
"outcome": "BLOCKED",
"stage": "ingress",
"artifactHash": null,
"validatorVersions": {
"pii-phi": "1.1.1",
"prompt-injection": "1.2.0",
"security-secrets": "1.1.0"
},
"scenario": {
"id": "block-input",
"mode": "synthetic-provider",
"providerCalls": 0,
"inputWasCustomized": false,
"ingress": {
"decision": "BLOCKED",
"outcome": "BLOCKED"
}
}
},
{
"id": "20000000-0000-4000-8000-000000000002",
"requestId": "30000000-0000-4000-8000-000000000002",
"decisionEventId": "40000000-0000-4000-8000-000000000002",
"createdAt": "2026-10-29T18:06:55.808Z",
"decision": "BLOCKED",
"overallScore": 18,
"tier": "HIGH",
"domainScores": [
{
"score": 18,
"domain": "SECURITY",
"maxScore": 18
}
],
"ruleIds": [
"PROMPT_INJECTION_HEURISTIC"
],
"validatorIds": [
"pii-phi",
"prompt-injection",
"security-secrets"
],
"remediationTypes": [
"BLOCK"
],
"timings": {
"totalMs": 1,
"upstreamMs": 0,
"preflightMs": 0,
"postflightMs": 1
},
"policyPackId": "developer-data-guard",
"policyPackVersion": "1.0.0",
"model": "lsas-evaluation-fixture-1",
"provider": "lsas-evaluation-recording",
"outcome": "BLOCKED",
"stage": "ingress",
"artifactHash": null,
"validatorVersions": {
"pii-phi": "1.1.1",
"prompt-injection": "1.2.0",
"security-secrets": "1.1.0"
},
"scenario": {
"id": "block-input",
"mode": "synthetic-provider",
"providerCalls": 0,
"inputWasCustomized": false,
"ingress": {
"decision": "BLOCKED",
"outcome": "BLOCKED"
}
}
},
{
"id": "20000000-0000-4000-8000-000000000003",
"requestId": "30000000-0000-4000-8000-000000000003",
"decisionEventId": "40000000-0000-4000-8000-000000000003",
"createdAt": "2026-10-27T18:47:17.044Z",
"decision": "ALLOW",
"overallScore": 0,
"tier": "LOW",
"domainScores": [],
"ruleIds": [],
"validatorIds": [
"pii-phi",
"prompt-injection",
"security-secrets"
],
"remediationTypes": [],
"timings": {
"totalMs": 2142,
"upstreamMs": 1763,
"preflightMs": 0,
"postflightMs": 0
},
"policyPackId": "developer-data-guard",
"policyPackVersion": "1.0.0",
"model": "lsas-evaluation-action-fixture-1",
"provider": "lsas-evaluation-action-recording",
"outcome": "RELEASED",
"stage": "egress",
"artifactHash": "e7b496a2f19ad8397da7867d92d20fd4f2971c89e841727372d10b5f109e4df9",
"action": {
"caseId": "allowed-read",
"mode": "synthetic-reader",
"authorization": "AUTHORIZED",
"reasonCodes": [],
"invocationCount": 1,
"contentDecision": "ALLOW",
"requestedSkillId": "lsas.evaluation.records.read",
"destination": "lsas-evaluation://records/approved",
"manifest": {
"id": "lsas.evaluation.records.read",
"version": "1.0.0",
"digest": "48e7ea84c5d0bc7e9c6674d6c13e58113785407b4b6070323450fb473e4a4e02"
},
"grant": {
"grantId": "50000000-0000-4000-8000-000000000003",
"inputDigest": "57917c57c9ced57600eef5fa07b6aaeb8ae2f5a6f50517ac5b3e824c953f6088",
"actionDigest": "b4ab0a61ba3051382b6028522386cfb23555d1728aec910d87f58bd0b750a0c3",
"authorityDigest": "99ef788ca86d0fb2dde030cca50ef0b83a756afe45c9e81340cfe7d58d939ee7",
"expiresAt": 1793126844312
},
"executionReceipt": {
"state": "computed-unreleased",
"outputDigest": "e7b496a2f19ad8397da7867d92d20fd4f2971c89e841727372d10b5f109e4df9"
}
},
"validatorVersions": {
"pii-phi": "1.1.1",
"prompt-injection": "1.2.0",
"security-secrets": "1.1.0"
}
},
{
"id": "20000000-0000-4000-8000-000000000004",
"requestId": "30000000-0000-4000-8000-000000000004",
"decisionEventId": "40000000-0000-4000-8000-000000000004",
"createdAt": "2026-10-27T18:46:48.707Z",
"decision": "ALLOW",
"overallScore": 0,
"tier": "LOW",
"domainScores": [],
"ruleIds": [],
"validatorIds": [
"pii-phi",
"prompt-injection",
"security-secrets"
],
"remediationTypes": [],
"timings": {
"totalMs": 380,
"upstreamMs": 377,
"preflightMs": 2,
"postflightMs": 1
},
"policyPackId": "developer-data-guard",
"policyPackVersion": "1.0.0",
"model": "lsas-evaluation-fixture-1",
"provider": "lsas-evaluation-recording",
"outcome": "RELEASED",
"stage": "egress",
"artifactHash": "1abae86b19980619984798c8944885dc062380ff03cf811cce82d8cf938141cb",
"validatorVersions": {
"pii-phi": "1.1.1",
"prompt-injection": "1.2.0",
"security-secrets": "1.1.0"
},
"scenario": {
"id": "safe-release",
"mode": "synthetic-provider",
"providerCalls": 1,
"inputWasCustomized": false,
"ingress": {
"decision": "ALLOW",
"outcome": "RELEASED",
"artifactHash": "7b1dc83bdc9c0dc68817dc653d99ef285b4e88cff992207c13cb19eb3b1d2bb2"
}
}
}
]
}