AccessGraph
Maps agent surfaces, actions, business-critical asset classes, and required controls in one structured record.
cg-accessgraph-1.0.0
See the access chain other scanners miss.
CapitalGuard AccessGraph connects each agent surface to what it can read, change, execute, or transfer, the business asset at risk, and the control that should stand between them.
CapitalGuard AccessGraph
Read. Change. Execute. Transfer.
Repository workspaceIllustrative coding-agent policy
ReadCredential-bearing pathsAccount and infrastructure accessBlock
ChangeAuth and billing logicRevenue and account integrityHuman approval
ExecuteLocal commandsDeveloper and repository stateHuman approval
TransferPrompts and logsPrivate context disclosureRedact
Licensed scans generate this map locally from aggregate exposure state without placing secret values in the graph.CGX-8D42-7AC1-55F0
One ownable security record
A scan becomes a baseline the company can keep.
The durable value is not a colorful score alone. It is the linked record of exposure, policy, change, and evidence that becomes more useful after every authorized review.
Guardprint
Fingerprints aggregate exposure state so a later scan can identify change without publishing repository contents.
Policy artifact
Turns the map into blocked paths, approval gates, redaction rules, and untrusted-instruction boundaries.
Verified history
Keeps score, graph version, policy version, delivery state, and approved verification attached to the review.
Multi-hop path analysis
See where one weak boundary can lead.
Exposure Graph extends the local AccessGraph into an explainable path: entry point, vulnerable component, crossed trust boundary, downstream assets, blast radius, required control, and retained regression test.
Open Attack LabCapitalGuard Exposure Graph
Synthetic repository. No customer data.
Inputs
RepositoryPrompt or instruction
Agent runtime
AgentModelAgent handoff
Authority
ToolMCP serverCredentialPermission boundary
Downstream
APIData storeExternal dependency
Repository instruction to release credential
Entry point
Untrusted instruction in repository documentation
Vulnerable component
Coding agent with command and workflow access
Trust boundary crossed
Repository text crosses into executable agent instruction
Downstream assets
CI identityRelease workflowPackage registry
A manipulated task could affect published code, customers, and the company release chain.
Required control
Treat repository text as untrusted, block credential paths, and require approval for workflow or release actions.
CGAL-INJ-0015 evidence signalsHigh confidence
Control language
Four controls teams can act on.
CapitalGuard turns exposure into an explicit boundary instead of pretending every risk can be fixed automatically.
Block
Deny access to material that should never enter agent context.
Human approval
Pause high-impact changes and execution until an owner reviews them.
Redact
Keep sensitive values out of prompts, logs, reports, and transfers.
Treat as untrusted
Prevent repository text from silently becoming operating instruction.
Privacy boundary
A fingerprint of state, not a vault of code.
The licensed local scanner generates aggregate category counts and a reproducible Guardprint. Source contents, raw secret values, customer identity, and repository name are not required in the AccessGraph output.
Privacy-safe by default
Aggregate asset-surface counts
Detected agent configuration categories
Installed control state
Versioned local Guardprint
Previous-to-current drift status
Start the baseline