richfrem/agent-plugins-skills / plugins/agent-agentic-os/.apm/skills/running-control-plane
running-control-plane
FlaggedCRITICAL INTAKE GATEWAY: use at the start of any non-trivial engineering task, feature, refactor, multi-file bugfix, work package or new idea, including requests to explore, brainstorm, build, fix or change something, before planning or code changes. Enforces read-only discovery, 1-3 structured questions with defaults, control-plane registration and the immutable four-pillar TASK_SPEC.md. Also runs the rest of the control plane: health ("run health check", "check os metrics", "system monitor", event bus, locks, memory, "is my signing identity ready" status check), transition simulation ("dry run transition", "simulate transition", "check edge X -> Y"), worktree selection, and gating on/off ("disable the control plane", "turn off the push guard", "stop gating my commits"). Not for project setup or key creation (use setting-up-agentic-os) or improving skills (use improving-skills-in-lab).
tests/test_control_plane_state_machine.py
View on GitHub"""
test_control_plane_state_machine.py — State-Machine Domain Extraction (issue-524)
=====================================================================================
Purpose:
Unit tests for control_plane/state_machine.py's StateMachine in complete isolation — no
SQLite, no ControlPlane — proving state-machine validation (known-state check + adjacency
legality) is genuinely independently testable domain logic. Added after external
post-implementation review (round 2) found this validation still embedded directly in
ControlPlane.transition(), contradicting the plan's declared "state-machine validation"
separation. Also includes an integration test proving ControlPlane.transition() actually
delegates to self._state_machine rather than reimplementing the checks inline.
Key Input Dependencies:
None — pure in-memory StateMachine instance.
Key Functions:
- test_validate_known_state_rejects_unknown_state()
- test_validate_known_state_accepts_every_canonical_state()
- test_validate_adjacency_rejects_illegal_edge()
- test_validate_adjacency_accepts_every_legal_edge()
- test_control_plane_transition_delegates_to_state_machine()
"""
import io
import sys
from pathlib import Path
import pytest
SCRIPTS_DIR = Path(__file__).resolve().parent.parent / "scripts"
if str(SCRIPTS_DIR) not in sys.path:
sys.path.insert(0, str(SCRIPTS_DIR))
from control_plane.state_machine import StateMachine, CANONICAL_STATES, ALLOWED_TRANSITIONS, InvalidStateTransition
from control_plane.registry import TransitionRegistry
from control_plane.constants import (
STATE_INTAKE, STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_MULTI_AGENT_REVIEW, STATE_PLAN_REVIEW, STATE_AWAITING_APPROVAL, STATE_APPROVED, STATE_IN_WORKTREE, STATE_WORKTREE_REVIEW, STATE_MULTI_AGENT_CODE_REVIEW, STATE_VERIFY_EXIT, STATE_RETROSPECTIVE, STATE_DONE, STATE_ROLLED_BACK, STATE_ESCALATED,
)
def test_validate_known_state_rejects_unknown_state():
sm = StateMachine()
with pytest.raises(InvalidStateTransition, match="Unknown state: NOT_A_REAL_STATE"):
sm.validate_known_state("NOT_A_REAL_STATE")
def test_validate_known_state_accepts_every_canonical_state():
sm = StateMachine()
for state in CANONICAL_STATES:
sm.validate_known_state(state) # must not raise
def test_validate_adjacency_rejects_illegal_edge():
# DONE -> INTAKE is now the deliberate reset_to_intake recovery edge (added
# for control-plane-reset-transition) — no longer illegal, so it can't be
# used as the illegal-edge example here. DONE -> WORKTREE_REVIEW remains
# illegal (DONE is still terminal except for the one explicit reset edge).
sm = StateMachine()
with pytest.raises(InvalidStateTransition, match="Cannot transition task 't1' from 'DONE' to 'WORKTREE_REVIEW'"):
sm.validate_adjacency("t1", STATE_DONE, STATE_WORKTREE_REVIEW)
def test_validate_adjacency_accepts_every_legal_edge():
sm = StateMachine()
for from_state, to_states in ALLOWED_TRANSITIONS.items():
for to_state in to_states:
sm.validate_adjacency("t1", from_state, to_state) # must not raise
def test_human_recovery_target_is_not_limited_by_normal_dag():
sm = StateMachine()
sm.validate_recovery_target("t1", STATE_DONE, STATE_IN_WORKTREE)
with pytest.raises(InvalidStateTransition, match="must differ"):
sm.validate_recovery_target("t1", STATE_DONE, STATE_DONE)
def test_human_recovery_from_done_to_worktree_requires_and_consumes_approval(tmp_path):
from agent_control import ControlPlane
cp = ControlPlane(db_path=tmp_path / "control_plane.db")
cp.create_task(task_id="recovery-done-1", title="Reopen", runtime_tool="claude")
cp.coordinate_transition("recovery-done-1", STATE_DONE, "human", "Close fixture", interactive=True) # signed close
token = cp.record_recovery_approval(
"recovery-done-1", STATE_IN_WORKTREE, "human", actor="human",
reason="Missed implementation fix requires rework",
)
record = cp.apply_recovery_transition(
"recovery-done-1", STATE_IN_WORKTREE, token, "human", "Reopen for bounded rework"
)
assert record.from_state == STATE_DONE
assert record.to_state == STATE_IN_WORKTREE
assert cp.get_task("recovery-done-1")["state"] == STATE_IN_WORKTREE
def test_every_non_done_state_has_a_signed_closure_edge():
for state in CANONICAL_STATES:
if state != STATE_DONE:
assert STATE_DONE in ALLOWED_TRANSITIONS[state]
@pytest.mark.no_auto_signer
def test_direct_agent_close_is_refused_without_an_openssh_signature(tmp_path):
"""auth-ciba-increment-b: closing a task is a cryptographic gate. An agent (or anyone) calling the facade
without a signed transition_request fails closed: the state does not move."""
from agent_control import ControlPlane
from control_plane.coordinator import HumanProofRequired
cp = ControlPlane(db_path=tmp_path / "control_plane.db")
cp.create_task(task_id="force-1", title="Close", runtime_tool="claude")
with pytest.raises((HumanProofRequired, Exception)) as excinfo:
cp.transition("force-1", STATE_DONE, "agent", "close now")
assert "signature" in str(excinfo.value).lower() or "HUMAN_PROOF_REQUIRED" in str(excinfo.value)
assert cp.get_task("force-1")["state"] == STATE_INTAKE
def test_the_force_close_bypass_parameters_no_longer_exist(tmp_path):
"""The force_close / human_authorization keyword bypass was removed; passing it is a TypeError, not a
quiet fallback, on both the coordinator and the ControlPlane facade."""
from agent_control import ControlPlane
from control_plane.coordinator import TransitionCoordinator
cp = ControlPlane(db_path=tmp_path / "control_plane.db")
cp.create_task("force-kw", "Close", "claude")
with pytest.raises(TypeError, match="force_close"):
TransitionCoordinator(cp).coordinate_transition(
"force-kw", STATE_DONE, "human", "close", force_close=True, human_authorization="FORCE_CLOSE", interactive=True,
)
with pytest.raises(TypeError, match="force_close"):
cp.coordinate_transition(
"force-kw", STATE_DONE, "human", "close", force_close=True, human_authorization="FORCE_CLOSE", interactive=True,
)
assert cp.get_task("force-kw")["state"] == STATE_INTAKE
@pytest.mark.no_auto_signer
@pytest.mark.parametrize("typed", ["FORCE_CLOSE", "FORCE_DONE", "YES", "1"])
def test_a_typed_word_at_an_interactive_prompt_cannot_close_a_task(tmp_path, typed):
"""Even a real interactive terminal and the old magic words do nothing without a signature: the coordinator
halts with HUMAN_PROOF_REQUIRED and the task does not move."""
from agent_control import ControlPlane
from control_plane.coordinator import HumanProofRequired, TransitionCoordinator
cp = ControlPlane(db_path=tmp_path / "control_plane.db")
cp.create_task("force-typed", "Close", "claude")
prompts = []
coord = TransitionCoordinator(cp, input_fn=lambda p: prompts.append(p) or typed, output_stream=io.StringIO())
with pytest.raises(HumanProofRequired):
coord.coordinate_transition("force-typed", STATE_DONE, "human", "close", interactive=True)
assert cp.get_task("force-typed")["state"] == STATE_INTAKE
assert prompts == [] # nothing was offered as an authorizing prompt
@pytest.mark.no_auto_signer
def test_a_noninteractive_human_actor_string_cannot_close_a_task(tmp_path):
from agent_control import ControlPlane
from control_plane.coordinator import HumanProofRequired
cp = ControlPlane(db_path=tmp_path / "control_plane.db")
cp.create_task("force-spoof", "Close", "claude")
with pytest.raises(HumanProofRequired):
cp.coordinate_transition("force-spoof", STATE_DONE, "human", "close", interactive=False)
assert cp.get_task("force-spoof")["state"] == STATE_INTAKE
def test_a_signed_close_persists_from_any_state(tmp_path):
"""The positive path: with the (test) human signing the real request, closure commits."""
from agent_control import ControlPlane
cp = ControlPlane(db_path=tmp_path / "control_plane.db")
cp.create_task(task_id="force-2", title="Close", runtime_tool="claude")
record = cp.coordinate_transition("force-2", STATE_DONE, "human", "Emergency close", interactive=True)
assert record.from_state == STATE_INTAKE
assert cp.get_task("force-2")["state"] == STATE_DONE
assert cp._persistence.get_last_transition("force-2").to_state == STATE_DONE
def test_every_non_done_state_exposes_a_proof_gated_closure_edge():
"""Closure is legal from every non-terminal state, and every such edge is cryptographic: it declares
requires_cryptographic_proof, authority type cryptographic_proof, and no FORCE_* accepted-answer question."""
registry = TransitionRegistry.load_default()
proof_edges = registry.proof_required_edges()
for state in CANONICAL_STATES:
if state == STATE_DONE:
continue
assert STATE_DONE in ALLOWED_TRANSITIONS[state]
template = registry.get_template(state, STATE_DONE)
assert template is not None
assert (state, STATE_DONE) in proof_edges
assert template.authority.get("type") == "cryptographic_proof"
for question in template.human_questions:
assert not {"FORCE_DONE", "FORCE_CLOSE"} & set(question.get("accepted_answers", []))
def test_signed_close_reaches_done_from_every_nonterminal_state(tmp_path):
"""Exercise every closure source edge through the coordinator, real SQLite and the real (test) signer."""
from agent_control import ControlPlane
from control_plane.coordinator import TransitionCoordinator
from helpers.gate1_fixtures import init_git_worktree
paths = {
STATE_INTAKE: [],
STATE_INTERVIEW: [STATE_INTERVIEW],
STATE_DRAFT_PLAN: [STATE_INTERVIEW, STATE_DRAFT_PLAN],
STATE_PLAN_REVIEW: [STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_PLAN_REVIEW],
STATE_MULTI_AGENT_REVIEW: [STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_PLAN_REVIEW, STATE_MULTI_AGENT_REVIEW],
STATE_AWAITING_APPROVAL: [STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_PLAN_REVIEW, STATE_AWAITING_APPROVAL],
STATE_APPROVED: [STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_PLAN_REVIEW, STATE_AWAITING_APPROVAL, STATE_APPROVED],
STATE_IN_WORKTREE: [STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_PLAN_REVIEW, STATE_AWAITING_APPROVAL, STATE_APPROVED, STATE_IN_WORKTREE],
STATE_WORKTREE_REVIEW: [STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_PLAN_REVIEW, STATE_AWAITING_APPROVAL, STATE_APPROVED, STATE_IN_WORKTREE, STATE_WORKTREE_REVIEW],
STATE_MULTI_AGENT_CODE_REVIEW: [STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_PLAN_REVIEW, STATE_AWAITING_APPROVAL, STATE_APPROVED, STATE_IN_WORKTREE, STATE_WORKTREE_REVIEW, STATE_MULTI_AGENT_CODE_REVIEW],
STATE_VERIFY_EXIT: [STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_PLAN_REVIEW, STATE_AWAITING_APPROVAL, STATE_APPROVED, STATE_IN_WORKTREE, STATE_WORKTREE_REVIEW, STATE_VERIFY_EXIT],
# RETROSPECTIVE is closed by retrospective_to_done, which adds retrospective_done_guard and the full-suite
# receipt on top of the signature; those guards are covered by test_transition_guidance / test_agent_control.
STATE_ROLLED_BACK: [STATE_INTERVIEW, STATE_DRAFT_PLAN, STATE_PLAN_REVIEW, STATE_AWAITING_APPROVAL, STATE_APPROVED, STATE_IN_WORKTREE, STATE_ROLLED_BACK],
STATE_ESCALATED: [STATE_ESCALATED],
}
registry = TransitionRegistry.load_default()
for state, path in paths.items():
# <repo>/context/control_plane.db makes tmp the repo root, so the dirty-main check does not inspect the real checkout
(tmp_path / state.lower() / "context").mkdir(parents=True)
control_plane = ControlPlane(db_path=tmp_path / state.lower() / "context" / "control_plane.db")
control_plane.init_db()
task_id = f"close-{state.lower()}"
control_plane.create_task(task_id, "Closure simulation", "simulator")
current = STATE_INTAKE
for next_state in path:
template = registry.get_template(current, next_state)
assert template is not None
source_transition = control_plane._persistence.get_last_transition(task_id)
question_ids = list(template.stage_question_ids or [])
question_ids += [question["question_id"] for question in template.human_questions]
if template.approval.get("required") and template.approval.get("approver_role", "human") == "human":
question_ids.append(f"approval_{template.transition_id}")
for question_id in question_ids:
control_plane.record_decision(
task_id, source_transition.transition_id, current, next_state,
question_id, "SIMULATED_HUMAN_CONFIRMATION", actor="human",
)
if (current, next_state) in registry.proof_required_edges():
# a cryptographic gate: setup walks it through the real signed flow, never a direct persistence write
if next_state == STATE_VERIFY_EXIT:
control_plane.update_worktree(task_id, str(init_git_worktree(tmp_path / f"wt-{task_id}")), f"sim/{task_id}", "written_in_worktree")
control_plane.coordinate_transition(task_id, next_state, "human", f"legal setup for {state}", interactive=True)
else:
assert control_plane._persistence.apply_transition(
task_id, current, next_state, "simulator", f"legal setup for {state}"
)
current = next_state
record = TransitionCoordinator(control_plane, registry=registry, output_stream=io.StringIO()).coordinate_transition(
task_id, STATE_DONE, "human", f"Signed close from {state}", interactive=True,
)
assert (record.from_state, record.to_state) == (state, STATE_DONE)
assert control_plane.get_task(task_id)["state"] == STATE_DONE
def test_control_plane_transition_delegates_to_state_machine(tmp_path, monkeypatch):
"""Integration: ControlPlane.transition() calls self._state_machine.validate_known_state()
and validate_adjacency() rather than checking CANONICAL_STATES/ALLOWED_TRANSITIONS inline —
proven by monkeypatching StateMachine to record calls and confirm they happened."""
from agent_control import ControlPlane
calls = []
original_validate_known = StateMachine.validate_known_state
original_validate_adjacency = StateMachine.validate_adjacency
def recording_validate_known(self, to_state):
calls.append(("validate_known_state", to_state))
return original_validate_known(self, to_state)
def recording_validate_adjacency(self, task_id, current_state, to_state):
calls.append(("validate_adjacency", task_id, current_state, to_state))
return original_validate_adjacency(self, task_id, current_state, to_state)
monkeypatch.setattr(StateMachine, "validate_known_state", recording_validate_known)
monkeypatch.setattr(StateMachine, "validate_adjacency", recording_validate_adjacency)
cp = ControlPlane(db_path=tmp_path / "control_plane.db")
cp.create_task(task_id="t1", title="State Machine Delegation Task", runtime_tool="claude")
cp.transition(task_id="t1", to_state=STATE_INTERVIEW, actor="user", reason="test")
assert ("validate_known_state", STATE_INTERVIEW) in calls
assert ("validate_adjacency", "t1", STATE_INTAKE, STATE_INTERVIEW) in calls
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 471k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |