Controls · scope and mapping
Every control in this product is paired below with the file or route that carries it and with the standard identifiers it speaks to. Each row carries a status word, not a green cell: evidences where the artefact is the control itself, maps to where it addresses the same concern without discharging it, and not assessed where nothing here speaks to the concern.
Scope and control matrix
1 evidences · 15 maps to · 0 not assessed · 16 controls
Nothing here is certified or independently assessed.
No certification is claimed and no attestation report exists. The mapping is the Security Laboratory's own proposal, and no auditor has reviewed it. A mapping row is a proposal, not an audit finding: it names the concern a control speaks to and stops where the control's own boundary sentence stops.
| Control | Artefact | ISO/IEC 27001:2022 Annex A | SOC2 common criteria | Status |
|---|---|---|---|---|
CSP with per-request nonce csp-nonce The proxy sets a per-request nonce policy on HTML responses; a policy constrains inline script and eval rather than removing every injection path. | proxy middlewareproxy.ts | A.8.26 · A.8.28 | CC6.1 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
SSRF-safe fetch safe-fetch The module resolves, classifies and pins the connection so a rebind after the check cannot redirect it; callers that use a raw fetch are not covered. | SSRF-safe fetch modulelib/security/safe-fetch.ts | A.8.20 · A.8.21 | CC6.6 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Validated redirect loop validated-redirects Each hop is re-checked against the address classes under a hop limit; it protects opted-in callers and is not a network egress firewall. | validated redirect looplib/security/validated-redirects.ts | A.8.20 | CC6.6 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Address classifier address-classes A decision function classifies IPv4 and IPv6 literals and resolved addresses; what counts as allowed remains its callers' policy. | address classifierlib/security/address-classes.ts | A.8.20 · A.8.21 | CC6.6 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Client token audit token-audit The scan reports where JWT-shaped values sit in browser storage at mount; it reads storage and does not remove a token or change the auth model. | client token audit modulelib/security/token-audit.ts | A.5.17 · A.8.5 | CC6.1 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Per-route rate limiting rate-limit Tier-based budgets count requests per identity and window across 426 route files; a distributed attacker with many identities is outside the model. | rate-limit middlewarelib/middleware/rate-limit.ts | A.8.6 · A.8.20 | CC6.1 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Zod request validation request-validation The schema validator is the control itself and is called on 162 route files; it shows the measure exists and runs, not that any processing activity meets the duty. | schema validation wrapperlib/validation.ts | A.8.26 · A.8.28 | CC6.1 | evidences the artefact is the control's own implementation and a matching obligation row exists |
Context egress ledger egress-ledger The chat route records which context segments left the process, with consent decisions and hashes; a model path that bypasses the route is invisible to it. | context egress ledgerlib/chat/context/egress-ledger.ts | A.8.12 · A.8.15 | CC6.7 · CC7.2 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Chat tool grants with revoke tool-grants Grants are scoped, revoked and reaped through explicit routes; enforcement holds only where the tool loop reads the grant. | chat grants route family/api/chat/grants | A.5.15 · A.5.18 · A.8.2 | CC6.1 · CC6.3 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Tool-confirmation tokens tool-confirm-hmac An HMAC over the held call binds the confirmation the user saw to the arguments that execute; it does not decide which tools need confirmation. | tool-confirmation route/api/chat/tool-confirm | A.8.24 · A.8.26 | CC6.1 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Sandbox-agent event stream sandbox-monitor The SSE stream reports step events from a sandboxed session; it observes a session and is not an enforcement boundary around it. | sandbox event stream route/api/sandbox-agents/[sessionId]/events | A.8.15 · A.8.16 | CC7.2 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Step telemetry step-telemetry Per-StepKind latency percentiles are recorded for flow steps; the data says nothing about whether a step was authorised. | step telemetry modulelib/observability/step-telemetry.ts | A.8.15 · A.8.16 | CC7.2 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Marketplace consent diff marketplace-consent The diff compares declared install permissions with the set a listing now requests; a diff is a prompt for a decision, so the install path must still consult it. | consent diff modulelib/platform/consent-diff.ts | A.5.19 · A.5.20 | CC9.2 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
MCP secrets door secrets-door Secret material reaches an MCP server only through an explicit route behind an off-by-default flag; enabling it is an operator decision. | MCP secrets route/api/mcp/secrets | A.5.17 · A.8.24 | CC6.1 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Dependency audit delta dependency-audit The script compares the current advisory set against a stored baseline; nothing schedules it, so it is a point-in-time instrument rather than a running control. | dependency audit scriptscripts/security-deps.mjs | A.5.21 · A.8.8 | CC7.1 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Findings rescan and triage findings-rescan The rescan diffs repository findings against the prior pass and records a triage decision; it witnesses a point-in-time scan, not runtime behaviour. | findings rescan scriptscripts/security-rescan.mjs | A.8.8 · A.8.16 | CC7.1 | maps to the artefact speaks to the same concern; the standard's control is wider than one module |
Obligation register
2 evidences · 6 maps to · 4 not assessed · 12 rows
The full row-by-row register is not repeated here: it lives on the Evidence page, and a second copy would be two registers that can drift apart. The duty view on the Programme page carries the same vocabulary.
Status words: evidences — the artefact is the control's own implementation and a matching obligation row exists; maps to — the artefact speaks to the same concern; the standard's control is wider than one module; not assessed — nothing in this repository speaks to it.
Tool Catalogue · what would be connected
A catalogue entry is a claim about a wire: what would be connected, which scopes and data classes a grant would touch, and when the claim was last checked. The rows are fixture claims, not a live inventory, and every row carries the boundary that stops it overclaiming.
Tool catalogue
21 tools · 7 with a wire · 3 wire-tested in the last 90 days
What this is not
The catalogue is not the tool plane. It records the glue and assurance a lab would need around a tool, and nothing more — no row says a product should replace another, and no row describes a deployment. Every row is a fixture claim; nothing on this page is fetched from a live source.
Wazuh
Manager 4.14.9
ECS 9.5.0 · 42 fields mapped · dropped: process.parent_process.command_line, user.domain
Event ingest and query only. This row does not claim a managed detection service, and no response action is wired.
Security Lab / integration fabric
This demo does not connect — no OAuth flow runs from this page.
Elasticsearch
Self-managed 9.5.6
ECS 9.5.0 · 57 fields mapped · dropped: none
Read-only index access through a scoped key. The lab keeps no copy of the events and this row does not present Elasticsearch as a managed SIEM service.
Security Lab / integration fabric
This demo does not connect — no OAuth flow runs from this page.
OpenSearch
Security Analytics 3.9.0
Sigma 2.1.0 · 12 fields mapped · dropped: logsource.service, detection.condition
The prepackaged rules and their ECS/OCSF aliases are documentation-level here. The automatic mapping review the project asks for has not been run against this lab's traces.
Security Lab / detection
Shuffle
Community 2.3.0
The documentation describes an automation engine. This row does not claim a tested action API, and the suite keeps approval authority rather than the engine.
Security Lab / integration fabric
Cortex
TheHive Project OSS 4.2.0
Analyser and responder run requests only. This row does not include the case-management store, and it does not authorise any responder against a live tenant.
Security Lab / integration fabric
This demo does not connect — no OAuth flow runs from this page.
CrowdStrike Falcon
Falcon Insight
Read telemetry is the only scope described. Containment would be a separate consent and is not part of this row.
Security Lab / integration fabric
Trivy
OSS
OCSF 1.9.0 · 18 fields mapped · dropped: vulnerability.exploitability
Scanner output read through an API only. This row does not claim prioritisation and carries no EUVD enrichment.
Security Lab / detection
OWASP Amass
OSS
No scan runs without proof of ownership for the domain, and this row does not include an authorisation workflow.
Security Lab / exposure
Prowler
OSS
Read-only role assumption per cloud account. This row builds no workload protection and holds no cross-account role.
Security Lab / integration fabric
Google Cloud Sensitive Data Protection
Cloud service
Classification signals only. This row gives no residency answer and stores no classifier output.
Security Lab / exposure
Keycloak
OSS
Identity events and realm configuration read only. This row does not reason about privilege paths, and an SSO login is not treated as identity telemetry.
Security Lab / authority
This demo does not connect — no OAuth flow runs from this page.
Thinkst Canary
Self-hosted
Alert API only. Decoy placement is network work this lab does not perform, and a decoy count is not coverage.
Security Lab / detection
Microsoft Defender for Office 365
Microsoft 365 E5
Mailbox read scopes are the highest-consent item in the catalogue. This row does not include DMARC posture, which is assessed separately.
Security Lab / integration fabric
Velociraptor
OSS 0.77.3
A fixture collection request over mutual TLS only. This row deploys no collector, keeps no custody chain, and retains no endpoint payload.
Security Lab / DFIR
This demo does not connect — no OAuth flow runs from this page.
TheHive
OSS 4.1.24
Frozen OSS edition: no supported connector is documented for this line. The row is listed for identity, not for Connect, and the stale chip stays visible.
Security Lab / DFIR
TheHive
Cloud service
No connector is shipped for this edition. Listing the row establishes a namespace, not a wire, and no date is claimed because no check has run.
Security Lab / DFIR
YARA
Classic
Maintenance-mode engine; the connector is withdrawn and no new rules are accepted for this edition. The old receipt is history, not a live wire.
Security Lab / DFIR
Restic
OSS
Job status and immutability flags only. This row attests no restore; a green backup job is not a tested recovery.
Security Lab / integration fabric
This demo does not connect — no OAuth flow runs from this page.
Tailscale
SaaS
Policy and audit-stream read only. This row inspects no traffic and claims no identity-aware access enforcement in this lab.
Security Lab / integration fabric
OpenCTI
Community 7.261103.0
Graph read through a documented MCP interface only. An MCP listing is a namespace claim, not a trust surface, and this row curates no feeds.
Security Lab / intelligence
MISP
OSS 2.5.50
STIX 2.1 · 23 fields mapped · dropped: confidence, x-misp-galaxy
Collection and object read only. The caller keeps sharing-group and TLP handling; this wire does not dedupe, score or vouch for feed quality.
Security Lab / intelligence
This demo does not connect — no OAuth flow runs from this page.