One of init.md’s key principles is conditional artifact generation: no output file is universally required. A software project, an AI agent, a server, and a research environment each produce different initialization artifacts because they have different operational needs. Generating the same bundle of files for every target regardless of what it actually needs is not initialization — it is noise. The procedure determines which outputs apply during the DETERMINE stage, based on the target’s type, purpose, current state, and the selected template. Only those outputs are created in the CREATE stage.Documentation Index
Fetch the complete documentation index at: https://mintlify.com/bitwikiorg/init.md/llms.txt
Use this file to discover all available pages before exploring further.
Each artifact in the vocabulary below remains conditional. Project-specific configuration files, indexes, schema definitions, and validation artifacts may also be produced when the target requires them — the vocabulary is not exhaustive.
The operational artifact vocabulary
These are the named artifacts that may be produced during initialization. Each one has a defined purpose and a condition under which it applies.INIT.md / init.md
The target-specific initialization procedure for this target. Documents how to initialize this specific project, service, or environment. Generated when a persistent, human- or machine-readable initialization record is useful for the target — particularly when the target will be initialized repeatedly, handed off, or operated in multiple environments.
AGENTS.md
Agent instructions for contributors and operators working with AI agents on this target. Describes behavioral guidelines, contribution rules, and interaction expectations. Generated only when agent operation applies to the target and no valid instruction file already exists.
SELF.md
Agent identity, role, capabilities, and operational boundaries for a specific agent. Defines who the agent is, what it is responsible for, and what it is not permitted to do. Generated for agent targets that require a stable identity and role definition.
USER.md
Operator context for correct interaction with the target or agent. Describes the expected operator, their goals, their level of technical familiarity, and how the target should adapt its behavior for them. Generated when operator-specific context materially affects how the target operates.
TOOLS.md
Tool permissions and invocation rules for agent or automated targets. Lists which tools are available, which are restricted, under what conditions each tool may be invoked, and what the agent must not do with them. Generated for targets that use tool-calling systems or automated execution environments.
STATE.md
Current condition or pending work externalized as a structured record. Documents the target’s live operational state — what is in progress, what is blocked, what decisions are pending. Generated when state must be made visible across sessions, handoffs, or operator changes.
PLAN.md
Actionable work remaining, structured as a task list or work plan. Documents what needs to happen next in order to advance the target toward its intended operational result. Generated when meaningful work remains after initialization and that work needs to be tracked explicitly.
SNAPSHOT.md
Point-in-time state record for a resumable environment or system. Captures the target’s state at a specific moment so that work can be resumed, reviewed, or restored later. Generated when the target is ephemeral, checkpoint-based, or likely to be paused and resumed.
README.md
Project documentation for human readers. Describes what the project is, how to set it up, and how to use it. Generated only when project documentation is missing or inadequate — not to replace an existing README that already covers the necessary ground.
PRD.md
Product requirements document establishing the target’s goals, scope, constraints, and success criteria. Generated only when product requirements need to be established — not generated automatically for every project.
.env.example
Environment variable documentation that lists required and optional variables with placeholder values but no real secrets. Generated when the target uses environment variables that operators need to know about and that are not already documented elsewhere.
SERVER.md
Server-specific state record documenting infrastructure targets: server identity, running services, configuration references, health status, and operational notes. Generated for server or infrastructure targets only.
RUNBOOK.md
Operations handoff material for humans taking over operation of a running system. Documents startup procedures, common failure modes, recovery steps, and operational contacts. Generated when the target is a running service or system being handed off to a new operator.
Existing work
When initialization encounters existing artifacts, the default behavior is to prefer adaptation over replacement. The procedure should:- preserve relevant existing files that are complete and valid
- update incomplete or outdated files when the required changes are clear and bounded
- avoid generating duplicates of files that already exist and serve the same purpose
- avoid overwriting substantial user-authored work without a clear, specific reason to do so
- identify conflicts between existing and proposed artifacts before making changes
- reuse existing project conventions — naming patterns, directory structure, tooling choices — where they are already established