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).
scripts/control_plane/gate_evidence.py
View on GitHub"""
control_plane/gate_evidence.py
==============================
Purpose:
Authenticate a real-work task's DONE before git lets its branch leave the machine. The push
guard must not trust stored task state alone: a completed simulation database copied or
symlinked into <repo>/context/control_plane.db looks DONE. This re-verifies, with ssh-keygen,
every signed gate the task committed, against the repository's own human trust anchor:
- <repo>/context/control_plane.db is a regular file (not a symlink), with one hard link,
classified as real work, and stamped 'real' inside the database;
- the task is DONE and has stored evidence for its signed gates, including the gate into DONE;
- each stored challenge names this task and that transition, its signature verifies against
<repo>/context/identity/allowed_signers, and the signing key is a human key (never a key
registered in <repo>/context/simulation/identity, whatever the production file calls it);
- every signed gate used the same key (one approver per pipeline).
Cross-platform (macOS, Linux, Windows): Python stdlib plus ssh-keygen (OpenSSH 8.1+).
Key Input Dependencies:
- <repo>/context/control_plane.db (tasks, signed_gate_evidence, control_plane_identity)
- <repo>/context/identity/allowed_signers (the human trust anchor)
- control_plane.approver_policy (classification, key roles), control_plane.ssh_signing (verification)
Key Functions:
- verify_branch_done() -- list every reason a branch's task may not be pushed (empty list: allowed)
- _database_file_problems(), _load_task_evidence(), _check_evidence_row() -- its three steps
Usage:
python3 scripts/verify_gate_evidence.py --repo-root <repo> --branch <branch>
"""
import sqlite3
from pathlib import Path
from typing import List, Optional, Tuple
from control_plane.approver_policy import (
CONTEXT_REAL,
ApproverPolicyError,
classify_path,
database_identity,
key_role,
registered_simulation_keys,
)
def _task_for_branch(conn: sqlite3.Connection, branch: str) -> Optional[tuple]:
"""Return (task_id, state) of the latest task registered for a branch, or None."""
return conn.execute(
"SELECT task_id, state FROM tasks WHERE worktree_branch = ? ORDER BY created_at DESC LIMIT 1", (branch,)
).fetchone()
def _challenge_names(challenge: bytes, task_id: str, from_state: str, to_state: str) -> bool:
"""Return True if a signed challenge names exactly this task and transition."""
lines = challenge.decode("utf-8", "replace").splitlines()
return f"task: {task_id}" in lines and f"transition: {from_state} -> {to_state}" in lines
# The production database file itself must be trustworthy before anything in it is read
def _database_file_problems(db: Path, anchor: Path) -> List[str]:
"""Return problems with the production database file and trust anchor (empty when sound)."""
if db.is_symlink():
return [f"{db} is a symlink; the production control plane must be a regular file."]
if not db.is_file():
return [f"{db} does not exist."]
if db.stat().st_nlink != 1:
return [f"{db} has {db.stat().st_nlink} hard links; the production control plane must have exactly one."]
if classify_path(db) != CONTEXT_REAL:
return [f"{db} is not classified as a real-work database."]
if not anchor.is_file():
return [f"The human trust anchor {anchor} is missing."]
return []
# Read the branch's task and its stored signed-gate evidence (read-only)
def _evidence_history_problems(conn: sqlite3.Connection, task_id: str, evidence: list) -> List[str]:
"""Match every proof-bearing history edge to one consumed, content-bound request."""
from control_plane.registry import TransitionRegistry
from control_plane.ssh_signing import ChallengeError, derive_challenge_from_row
registry = TransitionRegistry.load_default()
problems = []
expected_requests = set()
by_request = {row[0]: row for row in evidence}
history = conn.execute(
"SELECT transition_id, from_state, to_state FROM task_transitions "
"WHERE task_id = ? ORDER BY transition_id", (task_id,),
).fetchall()
occupancy = None
for transition_id, from_state, to_state in history:
template = registry.get_template(from_state, to_state)
if template and template.requires_cryptographic_proof:
requests = conn.execute(
"SELECT request_id, jti FROM transition_request WHERE task_id = ? "
"AND from_state = ? AND to_state = ? AND occupancy_id = ? AND status = 'CONSUMED'",
(task_id, from_state, to_state, occupancy),
).fetchall()
if len(requests) != 1:
problems.append(f"Transition {transition_id} lacks exactly one consumed signed evidence request.")
else:
request_id, jti = requests[0]
expected_requests.add(request_id)
row = by_request.get(request_id)
if row is None:
problems.append(f"Transition {transition_id} ({from_state} -> {to_state}) lacks signed gate evidence.")
elif row[1:3] != (from_state, to_state) or jti != f"sshsig:{row[5]}":
problems.append(f"Evidence {request_id} disagrees with its committed transition request.")
else:
try:
expected = derive_challenge_from_row(conn, request_id)
if bytes(row[3]) != expected:
problems.append(f"Evidence {request_id} challenge disagrees with its original request.")
except ChallengeError as exc:
problems.append(f"Evidence {request_id}: {exc}")
occupancy = transition_id
if set(by_request) != expected_requests:
problems.append(f"Task {task_id} has signed evidence unrelated to its proof-bearing transition history.")
return problems
def _load_task_evidence(db: Path, branch: str) -> Tuple[List[str], Optional[str], list]:
"""Return (problems, task_id, evidence rows) for the task registered on `branch`."""
conn = sqlite3.connect(f"file:{db}?mode=ro", uri=True)
try:
identity = database_identity(conn)
if identity != CONTEXT_REAL:
return [f"{db} is stamped '{identity}', not 'real': a simulation or unidentified database cannot unlock a push."], None, []
row = _task_for_branch(conn, branch)
if row is None:
return [f"No task is registered for branch {branch}."], None, []
task_id, state = row
if state != "DONE":
return [f"Task {task_id} is in state {state}, not DONE."], task_id, []
has_table = conn.execute("SELECT 1 FROM sqlite_master WHERE type = 'table' AND name = 'signed_gate_evidence'").fetchone()
evidence = conn.execute(
"SELECT request_id, from_state, to_state, challenge, signature, fingerprint FROM signed_gate_evidence "
"WHERE task_id = ? ORDER BY request_id", (task_id,),
).fetchall() if has_table else []
return _evidence_history_problems(conn, task_id, evidence), task_id, evidence
finally:
conn.close()
# Re-verify one stored gate against the human trust anchor; return (problem or None, verified key)
def _check_evidence_row(row: tuple, task_id: str, anchor: Path, simulation_keys: set) -> Tuple[Optional[str], Optional[str]]:
"""Return (problem, fingerprint) for one evidence row; problem is None when it is a valid human signature."""
from control_plane.ssh_signing import SigningError, verify_signature
request_id, from_state, to_state, challenge, signature, fingerprint = row
challenge, signature = bytes(challenge), bytes(signature)
if not _challenge_names(challenge, task_id, from_state, to_state):
return f"Evidence {request_id} does not name task {task_id} and {from_state} -> {to_state}.", None
try:
verified = verify_signature(challenge, signature, allowed_signers=anchor)
except SigningError as exc:
return f"Evidence {request_id} ({from_state} -> {to_state}) does not verify against {anchor}: {exc}", None
if verified.fingerprint != fingerprint:
return f"Evidence {request_id} was signed by {verified.fingerprint}, not the recorded {fingerprint}.", None
try:
role = key_role(verified.fingerprint, anchor)
except ApproverPolicyError as exc:
return f"Evidence {request_id}: {exc}", None
if role != "human" or verified.fingerprint in simulation_keys:
return f"Evidence {request_id} ({from_state} -> {to_state}) was signed by the agent's simulation key, not a human.", verified.fingerprint
return None, verified.fingerprint
# Every reason this branch's task may not be pushed; an empty list means allowed
def verify_branch_done(repo_root: Path, branch: str) -> List[str]:
"""Return the problems that block pushing `branch`; [] when its task's DONE is authenticated."""
repo_root = Path(repo_root)
db = repo_root / "context" / "control_plane.db"
anchor = repo_root / "context" / "identity" / "allowed_signers"
problems = _database_file_problems(db, anchor)
if problems:
return problems
problems, task_id, evidence = _load_task_evidence(db, branch)
if problems:
return problems
if not evidence:
return [f"Task {task_id} has no signed gate evidence; its DONE cannot be authenticated against the human trust anchor."]
if not any(row[2] == "DONE" for row in evidence):
return [f"Task {task_id} has no signed evidence for its transition into DONE."]
simulation_keys = registered_simulation_keys(db)
keys = set()
for row in evidence:
problem, key = _check_evidence_row(row, task_id, anchor, simulation_keys)
if problem:
problems.append(problem)
if key:
keys.add(key)
if len(keys) > 1:
problems.append(f"Task {task_id}'s gates were signed by {len(keys)} different keys; one approver per pipeline is required.")
return problems
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 |