Reference

Reference

Lookup-oriented technical detail for the Agent M harness — CLI flags, config schemas, detection rules, lint checks, and the shipped-work log. These pages document what is, exactly; the How-to pages cover how to do a task and Architecture covers how it's built.

Install & layout

• Installer CLI — flags, prerequisites, and the ownership table for install.sh / install.ps1.
• Migration tool — migrate-to-user-scope.{sh,ps1}, the 4-state matrix, and the .agentm-migrate-record.json schema.
• Repo layout — the top-level directory map and the adapter parity table.
• Compatibility — supported hosts, the OS matrix, and the adapter contract.

Detection & config

• Detection rules — the built-in rules and what each attaches a rationale to.
• Project config — the project.json enablement-block schema.
• Auto-orchestration config — every config key + default (toggles · thresholds · cooldowns) and the auto-orchestration-state.json shape.

Memory & vault

• AgentMemory context payload — the canonical paste-anywhere payload's sections.
• Vault lint checks — the read-only vault_lint.py checks: id · severity · what each checks · suggested-fix shape.
• Note relatedness signals — the signals + thresholds notes_link_discovery.py scores on.
• Vault write protocol — the lock path, content-hash CAS, atomic-writer, and the pin-offline / partitioning operator habits for N≥2 concurrent writers.
• Process seam — the read-only client a process calls instead of reaching into the engine: the three functions, their signatures, and degrade contracts.
• Storage seam — the verbs the engine calls instead of touching files: the seven-verb StorageBackend contract, the Locator/Info/Capabilities types, and the DeviceLocalBackend that implements them over plain markdown (V5-1 parts 1–2).

CI & shipped work

• CI gates — what each CI workflow proves and the script behind it.
• Completed features — the reverse-chronological log of shipped work.

See also

How-to · Architecture · Home