Imported from trydavidqix/Lumenva (.codex/AGENTS.md). Install upstream with npx skills add trydavidqix/Lumenva --skill .codex. Copyright stays with the author.

Codex adapter — DeskcommCRM

This file supplements the root AGENTS.md for Codex-specific behavior.

Authority

• Canonical doctrine: ../CLAUDE.md.
• Portable contract: ../AGENTS.md.
• Shared modular rules: ../.claude/rules/.
• Doctrine reconciliation matrix: ../docs/harness-doctrine-matrix.md.
• Codex repo skill: ../.agents/skills/DeskcommCRM/SKILL.md.

The Codex skill is a bridge, not an independent source of conventions. If it disagrees with CLAUDE.md, follow CLAUDE.md. If historical wording conflicts with current Spec/PRD/business-rule, use the repository precedence and the reconciliation matrix rather than reviving a frozen snapshot.

Local/private configuration

Keep user credentials, private MCPs and machine-specific settings in ~/.codex/config.toml or the appropriate local store. Do not commit secrets to this repository.

Multi-agent support

• explorer — read-only evidence gathering.
• reviewer — read-only correctness/security/regression review.
• docs-researcher — read-only API/documentation verification.

Use these roles for their narrow purpose; do not make them alternate sources of product doctrine.

Workflow

Before writing code:

1. read CLAUDE.md;
2. read the relevant shared rule(s) — API, audit, LGPD, WAHA and data-modeling are separate domains, not generic security notes;
3. inspect existing code/specs/business rules;
4. use the appropriate process skill/workflow;
5. verify with evidence before declaring completion.

Run pnpm harness:check after changing harness/instruction files.