# OSuite Runtime Exposure Report

- **Generated:** 2026-08-04T00:27:48.589Z
- **Workspace:** n/a
- **Exposure score:** 0/100
- **Exposure status:** not_started

## Executive Summary

Connect an agent and run one governed action to build the first runtime exposure map.

## AI Runtime Inventory

No AI runtime exposure surface has been discovered yet.

| Metric | Value |
| --- | --- |
| Agents | 0 |
| Runtime adapters | 19 |
| Coverage tier | managed |
| Active sessions | 4 |
| Systems touched | 0 |
| Governed actions | 0 |
| High-impact systems | 0 |

_No records in this section._

| Runtime lane | Status | Family | Active sessions | Total sessions |
| --- | --- | --- | --- | --- |
| Claude Connectors Directory | ready | remote_mcp | 0 | 0 |
| ChatGPT Apps Connector | ready | remote_mcp | 0 | 0 |
| Codex Remote Connector | ready | remote_mcp | 0 | 0 |
| Claude Desktop Extension | ready | framework_sdk | 0 | 0 |
| Claude Code Hooks | ready | tool_hook_runtime | 4 | 4 |
| Codex Plugin | ready | tool_hook_runtime | 0 | 0 |
| Codex Hooks | ready | tool_hook_runtime | 0 | 0 |
| Azure Foundry | ready | managed_agent_platform | 0 | 0 |
| Gemini Enterprise Agent Platform | planned | managed_agent_platform | 0 | 0 |
| Bedrock AgentCore | planned | managed_agent_platform | 0 | 0 |
| OpenAI Agents SDK | planned | framework_sdk | 0 | 0 |
| OpenAI API | planned | framework_sdk | 0 | 0 |
| OpenAI Frontier | planned | managed_agent_platform | 0 | 0 |
| xAI API | planned | framework_sdk | 0 | 0 |
| Grok Business | planned | managed_agent_platform | 0 | 0 |
| Grok Enterprise | planned | managed_agent_platform | 0 | 0 |
| ServiceNow AI Agent Fabric | planned | managed_agent_platform | 0 | 0 |
| Salesforce Agentforce | planned | managed_agent_platform | 0 | 0 |
| LangGraph Platform | planned | framework_sdk | 0 | 0 |

| Session | Agent | Runtime lane | Status | Signature | Posture |
| --- | --- | --- | --- | --- | --- |
| 00a631a5-ae38-47f0-8849-0b96f24ac8eb | codex-cli-prod-smoke | Claude Code Hooks | active | workspace_key_optional | enforce |
| d54e5524-7d98-4bf2-b483-7e01f020b629 | codex-cli-prod-smoke | Claude Code Hooks | active | workspace_key_optional | enforce |
| 671ab267-4845-47f9-b017-7d17e442c81a | codex-cli-prod-smoke | Claude Code Hooks | active | workspace_key_optional | enforce |
| f7a64600-7032-42ce-bcaa-c1e8a2978a4c | claude-hooks-demo | Claude Code Hooks | active | workspace_key_optional | enforce |

## Runtime Coverage

Runtime Coverage separates Reference adapters, Managed OSuite adapters, and Enterprise custom adapters so customers can see which action lanes are understood, supported, and production-governed. Current top active tier: Managed OSuite adapters.

- **Status:** active
- **Selected tier:** managed

| Metric | Value |
| --- | --- |
| Reference lanes | 0 |
| Managed lanes | 8 |
| Enterprise custom lanes | 11 |
| Connected coverage | 1/19 |
| Runtime sessions | 4 |
| Covered agents | 0 |

| Tier | Count | User path | Studio behavior |
| --- | --- | --- | --- |
| Reference adapters | 0 | Install osuite-cava-core, define or reuse parser packs, run local canonicalization and receipt checks. | Shown as reference coverage only; high-impact production lanes should not be marked fully governed from this tier alone. |
| Managed OSuite adapters | 8 | Use the one-line installer or connector flow; OSuite maintains parser updates and evidence quality. | Appears as Runtime Coverage with parser posture, evidence gaps, unsupported actions, and remediation paths. |
| Enterprise custom adapters | 11 | Map sample events into CAVA fields, validate coverage, version the adapter, and deploy through OSuite. | Appears as customer runtime coverage with field mapping, versioning, approval routing, and proof export. |

| Runtime lane | Tier | Parser posture | Status | Disclosure |
| --- | --- | --- | --- | --- |
| Claude Connectors Directory | Managed OSuite adapters | managed | ready | Managed pack: OSuite maintains parser updates, coverage diagnostics, proof export, and policy routing. |
| ChatGPT Apps Connector | Managed OSuite adapters | managed | ready | Managed pack: OSuite maintains parser updates, coverage diagnostics, proof export, and policy routing. |
| Codex Remote Connector | Managed OSuite adapters | managed | ready | Managed pack: OSuite maintains parser updates, coverage diagnostics, proof export, and policy routing. |
| Claude Desktop Extension | Managed OSuite adapters | managed | ready | Managed pack: OSuite maintains parser updates, coverage diagnostics, proof export, and policy routing. |
| Claude Code Hooks | Managed OSuite adapters | managed | connected | Managed pack: OSuite maintains parser updates, coverage diagnostics, proof export, and policy routing. |
| Codex Plugin | Managed OSuite adapters | managed | ready | Managed pack: OSuite maintains parser updates, coverage diagnostics, proof export, and policy routing. |
| Codex Hooks | Managed OSuite adapters | managed | ready | Managed pack: OSuite maintains parser updates, coverage diagnostics, proof export, and policy routing. |
| Azure Foundry | Enterprise custom adapters | enterprise_custom | ready | Enterprise custom pack: OSuite maps private runtime fields, customer schemas, and sensitive business systems. |
| Gemini Enterprise Agent Platform | Enterprise custom adapters | enterprise_custom | planned | Enterprise custom pack: OSuite maps private runtime fields, customer schemas, and sensitive business systems. |
| Bedrock AgentCore | Enterprise custom adapters | enterprise_custom | planned | Enterprise custom pack: OSuite maps private runtime fields, customer schemas, and sensitive business systems. |
| OpenAI Agents SDK | Managed OSuite adapters | managed | planned | Managed pack: OSuite maintains parser updates, coverage diagnostics, proof export, and policy routing. |
| OpenAI API | Enterprise custom adapters | enterprise_custom | planned | Enterprise custom pack: OSuite maps private runtime fields, customer schemas, and sensitive business systems. |

## Governance Capital

What survives if AI budgets, providers, or internal sponsors change: the controlled runtime inventory, action authority, approval leases, exposure history, and evidence record.

- **Product surface:** AI Deployment Survival
- **Score:** 16/100
- **Status:** exposed

| Dimension | Score | Status | Why it matters |
| --- | --- | --- | --- |
| Runtime control assets | 21/100 | exposed | If the organization cannot see the runtime estate, AI budget cuts or provider swaps turn into another discovery project. |
| Bounded authority | 0/100 | not_started | Trust is easier to preserve after an incident when approvals are scoped, time-bound, and replay-resistant. |
| Evidence durability | 0/100 | not_started | A pilot becomes defensible only when the team can prove what happened without reconstructing it manually. |
| Runtime portability | 92/100 | strong | Portability keeps OSuite from depending on a single agent vendor cycle. |
| Exposure resilience | 0/100 | not_started | When AI hype corrects, teams keep the systems that can show controlled blast radius and remediation progress. |
| Policy-to-runtime binding | 0/100 | not_started | Policy language becomes capital only when it changes runtime behavior and leaves evidence. |

| Survival test | Posture | Explanation |
| --- | --- | --- |
| Budget correction | needs proof | Reduce proof gaps and observe-only lanes before using OSuite as a budget-defense artifact. |
| Provider churn | portable | The governance object can survive a runtime or model-provider change. |
| Incident review | fragile | Close evidence durability and approval binding before relying on post-incident reconstruction. |
| Procurement diligence | needs hardening | Map policy profile, runtime decisions, and proof closure into one exportable packet. |

## Top Runtime Exposures

No exposure backlog item is currently active.

_No records in this section._

## Vendor and Runtime Dependency Risk

3 dependency lane(s) are visible across providers, runtimes, tools, systems, approvals, and evidence.

| Lane | Count | Risk | Examples |
| --- | --- | --- | --- |
| Model and agent providers | 5 | distributed | OpenAI / Codex, Anthropic / Claude, Microsoft / Azure, LangGraph, MCP tool servers |
| Runtime adapters | 19 | controlled | Claude Connectors Directory, ChatGPT Apps Connector, Codex Remote Connector, Claude Desktop Extension, Claude Code Hooks, Codex Plugin |
| MCP and tool servers | 3 | review | Claude Connectors Directory, ChatGPT Apps Connector, Codex Remote Connector |
| External and business systems | 0 | not_started |  |
| Approval and authority lanes | 0 | not_started | PCAA, policy profile, Action Gate Lease |
| Evidence and receipts | 0 | not_started | CAVA receipt, PCAA proof bundle, decision record |

## Approval Lease Posture

No governed action has reached the runtime firewall yet.

| Lane | Count | Meaning |
| --- | --- | --- |
| Allow with proof | 0 | Action may continue when CAVA meaning, policy profile, and proof closure agree. |
| Ask for approval | 0 | Execution waits for a PCAA-recognized authority instead of trusting the agent runtime alone. |
| Block or fail closed | 0 | High-risk allow paths are treated as firewall defects until a bounded approval exists. |
| Observe only | 0 | OSuite can record evidence but cannot yet enforce before execution on this lane. |
| Expired or unbound | 0 | Unsigned actions or proof gaps cannot be reused as durable approvals. |

- **Lease-ready actions:** 0
- **Replay checks:** 0
- **Rejected replay checks:** 0

## External Verifier Coverage

No external verifier checkpoint has been attached to recent runtime actions yet.

- **Coverage:** 0%
- **Expected source class:** independent_mediator
- **Independent mediator references:** 0
- **Verified references:** 0
- **Valid but mapping pending:** 0
- **Execution divergence:** 0

_No records in this section._

## Framework Mapping

Runtime exposure evidence can be mapped into security and governance operating models.

| Framework | Fit |
| --- | --- |
| CTEM | scope, discover, prioritize, validate, and mobilize agent action exposures continuously. |
| AI-SPM | discover agents, runtime adapters, MCP/tool surfaces, reachable systems, and policy gaps. |
| AI runtime protection | turn prompt/tool/runtime events into pre-execution allow, ask, block, or observe lanes. |
| NIST AI RMF | map runtime evidence into Govern, Map, Measure, and Manage outputs. |
| OWASP Agentic AI | surface tool, authorization, identity, data, and autonomy risks as action-level controls. |
| MITRE ATLAS | preserve incident evidence for AI-specific attack paths and control validation. |

## Recommended Remediation

Run at least one governed action before remediation can be recommended.

_No records in this section._