| Format | Purpose | Repository | Tier |
|---|---|---|---|
META.a2ml |
Architecture decisions, governance |
hyperpolymath/meta-a2ml |
Primary |
STATE.a2ml |
Project state, milestones |
(companion to META) |
Primary |
ECOSYSTEM.a2ml |
Ecosystem positioning |
(companion to META) |
Primary |
PLAYBOOK.a2ml |
Executable plans, runbooks |
hyperpolymath/playbook-a2ml |
Execution |
AGENTIC.a2ml |
AI agent gating, safety |
hyperpolymath/agentic-a2ml |
Execution |
NEUROSYM.a2ml |
Symbolic semantics |
hyperpolymath/neurosym-a2ml |
Execution |
ANCHOR.a2ml |
Project recalibration |
hyperpolymath/anchor-a2ml |
Intervention |
PLAYBOOK.a2ml defines how permitted actions are executed, not whether they are permitted (that’s META) or whether they may proceed now (that’s AGENTIC).
PLAYBOOKs are derived from:
-
META rules (architectural decisions)
-
STATE context (current project phase)
-
Validated user intent (explicit present request)
# SPDX-License-Identifier: PMPL-1.0-or-later
[playbook.derivation-source]
type = "derived"
meta-rules = ["adr-001", "adr-015"]
state-context = "beta-phase"
timestamp = "2026-01-03T12:00:00Z"
[playbook.procedures.build]
description = "Build the project"
on-failure = "abort-and-notify"
[[playbook.procedures.build.steps]]
step = 1
action = "deno task build"
timeout = 300
[[playbook.procedures.build.steps]]
step = 2
action = "deno task test"
timeout = 600
[playbook.procedures.deploy]
description = "Deploy to production"
requires-confirmation = true
on-failure = "rollback-and-alert"
[[playbook.procedures.deploy.steps]]
step = 1
action = "backup-current"
[[playbook.procedures.deploy.steps]]
step = 2
action = "push-to-prod"
[[playbook.procedures.deploy.steps]]
step = 3
action = "verify-deployment"| Feature | Benefit |
|---|---|
Derived Authority |
Plans inherit authority from META, cannot introduce new permissions |
Gated Execution |
All execution is subject to AGENTIC safety controls |
Auditable |
All executions and failures are recorded in journals |
Machine-Readable |
Tooling can parse, validate, and execute plans |
Rollback-Aware |
Failure handling and rollback strategies built-in |
Cross-Format |
Integrates with META, STATE, AGENTIC, NEUROSYM |
PLAYBOOKs execute within a defined order:
1. META validation -> Constitutional authority check
2. AGENTIC gating -> Operational safety check
3. NEUROSYM semantics -> Proof obligations discharge
4. PLAYBOOK derivation -> Build executable plan
5. Execution -> Run the plan
6. ECOSYSTEM check -> Integrity verification
7. STATE update -> Record success/failureDeclares where the PLAYBOOK’s authority comes from:
-
type- "derived" (default) or "non-derived" (if META permits) -
meta-rules- Which ADRs authorize this PLAYBOOK -
state-context- Current project phase -
agentic-gate- AGENTIC status at derivation time
Executable plans with:
-
description- Human-readable purpose -
preconditions- Required state before execution -
steps- Ordered actions with timeouts -
postconditions- Expected state after success -
on-failure- Failure handling strategy
Notification configuration:
-
Severity levels (info, medium, high, critical)
-
Channels (slack, pager, email)
-
Escalation delays
# SPDX-License-Identifier: PMPL-1.0-or-later
# SPDX-FileCopyrightText: 2026 Your Name
# PLAYBOOK.a2ml -- Operational Runbook
[playbook.derivation-source]
type = "derived"
meta-rules = ["adr-001"]
timestamp = "2026-01-03T00:00:00Z"
[playbook.procedures.build]
description = "Build the project"
on-failure = "abort"
[[playbook.procedures.build.steps]]
step = 1
action = "deno task build"The formal specification is in the spec/ directory:
-
PLAYBOOK-FORMAT-SPEC.adoc - Complete specification
See the examples/ directory:
-
minimal.scm - Minimal valid PLAYBOOK
-
comprehensive.scm - Full-featured example
A PLAYBOOK is conformant if:
-
It is derived from META (or explicitly permitted as non-derived)
-
It does not bypass META or AGENTIC
-
Execution ordering is preserved
-
Failures are properly recorded
-
MIME Type:
application/vnd.playbook+a2ml -
File Extension:
.a2ml(conventionallyPLAYBOOK.a2ml) -
Encoding: UTF-8
If PLAYBOOK execution fails:
-
The system MUST rollback or branch
-
STATE MUST NOT be committed
-
The failure MUST be recorded in the journal
-
Alerts MUST be triggered
-
META.a2ml - Architecture decisions
-
AGENTIC.a2ml - Operational gating
-
NEUROSYM.a2ml - Symbolic semantics