An open standard for defining which decisions require an authorized human in AI-assisted systems.
AI proposes. Humans decide.
DECIDERS.md is a Markdown file placed at the root of a repository that declares the human decision-rights policy for AI-assisted systems. It specifies which decisions must remain with an authorized human, which actions an AI may take autonomously, and who holds each decision right.
Think of it as the governance complement to AGENTS.md. Where AGENTS.md tells agents how to work, DECIDERS.md tells the organization where human authority begins.
DECIDERS.md classifies every AI action into one of three tiers based on who may authorize it.
Routine, low-risk, reversible actions the AI may take without approval. Running tests, formatting code, generating documentation.
Consequential actions the AI may propose but cannot execute until a designated human decider approves. Merging to main, modifying infrastructure, sending external communications.
Actions that must be initiated and executed by a human. Production deployments, access control changes, financial approvals, compliance configuration.
Create a DECIDERS.md file at the root of your repository.
# DECIDERS.md ## Governance Statement This project uses AI agents for development and operations. All consequential decisions require approval from an authorized human decider. Routine, low-risk, reversible actions may be performed autonomously within the boundaries defined below. ## Decision Categories ### Tier 1 — Autonomous - Running test suites and linters - Creating and updating draft pull requests - Generating documentation from code - Formatting and refactoring within established patterns ### Tier 2 — Propose and Approve - Merging to main Decider: Tech Lead - Modifying CI/CD config Decider: DevOps Lead - Updating dependencies Decider: Tech Lead - External notifications Decider: Product Owner ### Tier 3 — Human Only - Production deployments Owner: VP Engineering - Access control changes Owner: Security Lead - Financial approvals Owner: CFO ## Escalation Mechanics Tier 2 requests posted to #approvals with @mention. Timeout: 4 hours business, 12 hours outside. Fallback: system holds until a human responds. ## Delegation Rules Deciders may delegate to a named backup for a defined period. Tier 3 decisions may not be delegated to AI systems.
The AI convention ecosystem has organized into distinct layers. DECIDERS.md fills the governance layer that no other convention addresses.
Create a DECIDERS.md file at the root of your repository, alongside your AGENTS.md.
Classify actions into autonomous, propose-and-approve, and human-only categories based on risk and reversibility.
Assign each consequential decision to a specific, accountable human. Not a team, not a role, not an algorithm.
Define how approval requests are routed, what timeouts apply, and what the system does when a decider is unreachable.
No. Tier 1 actions are explicitly autonomous. DECIDERS.md defines the boundary between autonomous and human-authorized actions. It doesn't require human approval for everything.
No. DECIDERS.md is standard Markdown. The three-tier structure and decider registry are recommended but not enforced. Adapt it to your governance needs.
They're complementary. AGENTS.md tells agents how to work in your project. DECIDERS.md tells the organization which decisions require human authority. One is operational. The other is governance.
Absolutely. Treat DECIDERS.md as a living governance document. Review it when team structure changes, when AI capabilities expand, or at regular governance intervals.
DECIDERS.md is an open standard published under the MIT license, offered as a complement to AGENTS.md. Its escalation mechanics were informed by publicly documented human-in-the-loop safety primitives — the THROTTLE, ESCALATE, and KILLSWITCH.md patterns. Contributions and critique are welcome on GitHub.