Answer: Add review at the tool-execution boundary: check sensitive proposed actions against approved systems, actions, and time limits before execution, route ambiguous or high-risk actions to a person, enforce independent filesystem and network restrictions, keep audit logs, and fail closed if the reviewer or policy is unavailable. Codex Auto-review doesn't automatically protect custom tools or external harnesses, so use the Guardrails and human review guidance for the Agents SDK pattern and the open-source reviewer policy as a reference. Status: Confirmed Entry: tool/codex/agents-and-automation/agent-workflow-guardrails Revision: 1 Digest: sha256:669401439f1db6d9 (changes whenever the content changes) Timeline: Created: 2026-11-13T03:48:40Z Last changed: 2026-11-13T03:48:40Z Last verified: 2026-11-20T13:36:22Z Next check: 2026-11-21T13:36:22Z Check interval: Daily Expected change: No fixed schedule Data: Field | Value | Source | As of Where to add review | tool-execution boundary | [1] | 2026-11-20 Pre-execution check | approved systems, actions, and time limits | [1] | 2026-11-20 High-risk routing | route ambiguous or high-risk actions to a person | [1] | 2026-11-20 Restrictions | independent filesystem and network restrictions | [1] | 2026-11-20 Failure behavior | fail closed if the reviewer or policy is unavailable | [1] | 2026-11-20 Auto-review limitation | Codex Auto-review doesn't automatically protect custom tools or external harnesses | [1] | 2026-11-20 Sources: [1] OpenAI. "Recommended configuration". https://learn.chatgpt.com/docs/cyber-safety/recommended-configuration.md Role: primary. Accessed: 2026-11-20T13:36:22Z Quote: "add review at the tool-execution boundary" (Supports: Where to add review) Quote: "Check sensitive proposed actions against the approved systems, actions, and time limits before execution" (Supports: Pre-execution check) Quote: "route ambiguous or high-risk actions to a person" (Supports: High-risk routing) Quote: "enforce independent filesystem and network restrictions" (Supports: Restrictions) Quote: "fail closed if the reviewer or policy is unavailable" (Supports: Failure behavior) Quote: "Codex Auto-review doesn't automatically protect custom tools or external harnesses" (Supports: Auto-review limitation) Snapshot SHA-256: 23c191226fa7ce5a0f5dd0667890a756be83752eee2e4586fde4e386eb0b257d (captured 2026-11-13T03:48:38Z) Verification: Method: Automated extraction from [1] with deepseek-flash (JSON output, thinking disabled). Every value was checked by code to appear verbatim in a quoted passage of [1]. On this check the text of [1] was unchanged since 2026-11-13T03:48:40Z, so the values verified then were kept. Cross-check: not-available at 2026-11-20T13:36:22Z. No independent second source is configured for this entry Last successful check: 2026-11-20T13:36:22Z Last attempt: 2026-11-20T13:36:22Z Notes and limitations: Extracted automatically from a single official source (section "Add guardrails to custom agent workflows"). Related: https://latestn.com/tool/codex/agents-and-automation/automate-security-reviews-ci-cd How do I automate Codex security reviews in CI/CD? https://latestn.com/tool/codex/agents-and-automation/background-hooks How do I run a Codex hook in the background? https://latestn.com/tool/codex/agents-and-automation/build-skill How do I build a skill for ChatGPT or Codex? https://latestn.com/tool/codex/admin-and-enterprise/restrict-sign-in-to-company-workspace How do I restrict Codex sign-in to my company workspace? https://latestn.com/tool/codex/admin-and-enterprise/review-codex-analytics How do I review Codex analytics for my workspace? https://latestn.com/tool/codex/admin-and-enterprise/set-global-baseline-environment-overrides How do I set a Global baseline policy and add environment overrides in Codex Agent Security? https://latestn.com/tool/ All entries in AI tools Change log: Time | Event | Change | Source 2026-11-13T03:48:40Z | added | Entry created | — Changes for this entry as a feed: https://latestn.com/tool/codex/agents-and-automation/agent-workflow-guardrails/latest Challenge this entry: Open challenges: None How: If a value here is wrong or out of date, send what is wrong and a link to evidence. Challenges are queued and checked against the evidence; they do not change the entry by themselves. If the entry changes as a result, the change log says so. Form: https://latestn.com/dispute/tool/codex/agents-and-automation/agent-workflow-guardrails POST: https://latestn.com/dispute/tool/codex/agents-and-automation/agent-workflow-guardrails with claim, evidence_url, and optionally field and contact (form-encoded or JSON) GET: https://latestn.com/dispute/tool/codex/agents-and-automation/agent-workflow-guardrails?claim={what is wrong}&evidence_url={link} How to cite: LatestN. (2026). How do I add guardrails to custom agent workflows in Codex? Revision 1, last verified 2026-11-20T13:36:22Z. https://latestn.com/tool/codex/agents-and-automation/agent-workflow-guardrails (accessed 2026-11-21T10:03:19Z). @misc{LatestN_tool_codex_agents_and_automation_agent_workflow_guardrails_r1, title = {How do I add guardrails to custom agent workflows in Codex?}, author = {{LatestN}}, year = {2026}, howpublished = {\url{https://latestn.com/tool/codex/agents-and-automation/agent-workflow-guardrails}}, note = {Revision 1; digest sha256:669401439f1db6d9; last verified 2026-11-20T13:36:22Z; accessed 2026-11-21T10:03:19Z} } Reuse: License: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/) Attribution: Copying and adapting are allowed, including commercially, if you credit LatestN and link to the entry you used. Each entry ends with a ready-made citation. URL: https://latestn.com/tool/codex/agents-and-automation/agent-workflow-guardrails