Screenata is now available in the Codex marketplace. You can ask your coding agent how your SOC 2 preparation is going, investigate a failing security check, and open the evidence behind it. [Install the Screenata plugin](https://chatgpt.com/plugins/plugin_asdk_app_6abf1aa2c6a0819190e085c4e13201ac). We built this for founders and engineers who handle compliance alongside product work. When an enterprise customer asks for a SOC 2 report, someone needs to work out what's missing, make the changes, and retain the proof. Screenata brings that work into the conversation where you're already building the product. The plugin connects Codex to your Screenata workspace through the Model Context Protocol, or MCP. It includes a compliance skill that guides the agent through investigation, approval, and verification. Your controls, policies, tests, and evidence remain in Screenata, with browser links for review. ## Start with a question about your audit After connecting your account and selecting a workspace, try: > How's our SOC 2 readiness? Show the audit cycle and recommend one next action. Screenata returns the cycle's readiness score and the underlying program information, including tests, policies, evidence gaps, and open risks. Codex can use those results to explain what needs attention and link you to the audit cycle. You can then narrow the question: > Which automated checks last failed? Show a few, including when they ran and whether we can act on them. A saved failure needs context. Its integration may have disconnected, its evidence may be old, or the test may now be marked not applicable. The agent can inspect the test before recommending a change. ## Follow a finding to its evidence We exercised this read-only workflow through the plugin while preparing this post. The connector returned audit-cycle readiness, a prioritized next action, and a page of saved failing checks. Opening one branch-protection test returned its last run, the flagged settings, linked policy claims, and evidence file references. It also reported an integration blocker. That meant the saved finding was available to investigate, but an exact repair target was unavailable. We stopped at investigation. No repository settings changed, and we did not claim a fresh passing result. For your own workspace, select a returned test and ask: > Explain this test's last failure, show the evidence, and tell me what needs to happen before we can verify it again. The useful output is specific: which setting failed, when the check observed it, what policy it supports, and what prevents the next step. You can follow the test's browser link to review the record yourself. ## Review a change, then verify it For supported checks with the necessary connection and permissions, Screenata can prepare a remediation plan. Other fixes require the coding agent to propose a local change, or a person to complete a step in the provider's console. Ask to see the affected resources, proposed changes, side effects, and rollback approach before approving a fix. Once the approved change is complete, rerun the check and inspect its result. Fixing one resource may leave other failures open, and the audit-cycle readiness score may stay the same. Evidence stays attached to the relevant test. Native check runs retain their proof. You can also submit redacted text evidence through the plugin; screenshots and PDFs go through the Screenata portal or CLI. ## Policies and other frameworks The plugin also supports policy drafting from answers you provide and record. Unanswered questions remain visible, and drafts require human review and approval. Alongside SOC 2 Type I and Type II, the plugin supports readiness work for HIPAA, ISO 27001, and ISO 42001. Infrastructure baseline scores and audit readiness measure different things. Screenata helps prepare the program and evidence; your auditor issues the report or certification. ## Try it in your workspace Open the [Screenata listing](https://chatgpt.com/plugins/plugin_asdk_app_6abf1aa2c6a0819190e085c4e13201ac), install the plugin, and connect your Screenata account. New users create their account in the browser. Select the workspace you want Codex to use, then start with the readiness question above. Cloud assessments require consent for the selected account. Provider credentials stay out of chat and tool arguments, and local repository assessment keeps source code on your machine. Start by finding one thing that needs attention. Open its evidence, understand the gap, and decide what to do next.