Compare commits

...
7 changed files with 1312 additions and 0 deletions
@@ -0,0 +1,46 @@
[Unit]
Description=AIOS generic governed Online Agent Worker (2026-08-15)
After=aios-agent-task-api.service aios-online-agent-worker-register-20260815.service
Wants=aios-online-agent-worker-register-20260815.service
[Service]
Type=simple
User=aios-engineering-worker
Group=aios-engineering-worker
SupplementaryGroups=gitops
LoadCredentialEncrypted=agent_token:/etc/credstore.encrypted/aios-online-agent-worker-20260815/agent_token
Environment=PYTHONUNBUFFERED=1
Environment=AIOS_AGENT_TASK_API_URL=http://127.0.0.1:18940
Environment=AIOS_GOVERNED_WRITE_API_URL=http://127.0.0.1:18892
Environment=AIOS_AGENT_ID=online-engineering-worker-cn-network-mkt-generic-20260815
Environment=AIOS_AGENT_TYPE=trusted-online-engineering-worker
Environment=AIOS_AGENT_VERSION=2026-08-15.3-generic
Environment=AIOS_AUTHORITY_GIT_DIR=/data/aios-gitops/repo.git
Environment=AIOS_WORKER_STATE_ROOT=/var/lib/aios-online-agent-worker-20260815
Environment=AIOS_WORKER_WORKSPACE_ROOT=/var/lib/aios-online-agent-worker-20260815/workspaces
Environment=AIOS_WORKER_POLL_SECONDS=5
Environment=AIOS_WORKER_HEARTBEAT_SECONDS=10
Environment=AIOS_WORKER_LEASE_SECONDS=300
ExecStart=/usr/bin/python3 -B /usr/local/lib/aios-online-agent-worker-20260815/online_agent_worker.py
Restart=always
RestartSec=5
TimeoutStopSec=15
NoNewPrivileges=true
PrivateTmp=true
ProtectSystem=strict
ProtectHome=true
ProtectKernelTunables=true
ProtectKernelModules=true
ProtectControlGroups=true
LockPersonality=true
RestrictSUIDSGID=true
RestrictRealtime=true
RestrictNamespaces=true
MemoryDenyWriteExecute=true
RestrictAddressFamilies=AF_UNIX AF_INET AF_INET6
ReadOnlyPaths=/data/aios-gitops/repo.git
ReadWritePaths=/var/lib/aios-online-agent-worker-20260815
UMask=0027
[Install]
WantedBy=multi-user.target
@@ -0,0 +1,36 @@
[Unit]
Description=AIOS generic Online Agent register-or-recover (2026-08-15)
After=network-online.target aios-gitops-model-api.service aios-agent-task-api.service
Wants=network-online.target
Before=aios-online-agent-worker-20260815.service
[Service]
Type=oneshot
RemainAfterExit=yes
User=root
Group=root
LoadCredential=bridge_control_token:/run/credentials/aios-gitops-model-api.service/agent_task_control_token
Environment=AIOS_AGENT_TASK_API_URL=http://127.0.0.1:18940
Environment=AIOS_AGENT_ID=online-engineering-worker-cn-network-mkt-generic-20260815
Environment=AIOS_AGENT_TYPE=trusted-online-engineering-worker
Environment=AIOS_AGENT_VERSION=2026-08-15.3-generic
Environment=AIOS_WORKER_ENCRYPTED_CREDENTIAL=/etc/credstore.encrypted/aios-online-agent-worker-20260815/agent_token
Environment=AIOS_WORKER_REGISTRATION_STATE=/var/lib/aios-online-agent-worker-20260815/registration
ExecStart=/usr/bin/python3 -B /usr/local/lib/aios-online-agent-worker-20260815/register.py
NoNewPrivileges=true
PrivateTmp=true
ProtectSystem=strict
ProtectHome=true
ProtectKernelTunables=true
ProtectKernelModules=true
ProtectControlGroups=true
LockPersonality=true
RestrictSUIDSGID=true
RestrictRealtime=true
RestrictAddressFamilies=AF_UNIX AF_INET AF_INET6
ReadWritePaths=/etc/credstore.encrypted /var/lib/aios-online-agent-worker-20260815
UMask=0077
TimeoutStartSec=120
[Install]
WantedBy=multi-user.target
@@ -0,0 +1,54 @@
# Generic Governed Online Agent Worker
This worker is a pull-based, restartable consumer for `aios.agent-task.v1`.
The service configuration contains only stable runtime identity, capability
names, API endpoints, heartbeat timing, and the credential projection. It does
not contain a task id, repository base OID, target branch, or plan-file path.
## Dynamic task contract
Engineering tasks carry `metadata.execution` with schema
`aios.online-agent.execution.v2`. The payload supplies the repository, exact
base OID, target branch, detached-worktree policy, adapter name, explicit file
change set, verification commands, and governed commit idempotency keys.
Task prose is descriptive only and is never passed to a shell.
The worker accepts only structured operations:
- `deterministic`: allow-listed file changes and test commands;
- `codex`: fail-closed until a task supplies an explicit Codex file/command
policy;
- `hermes`: fail-closed because the current Hermes runtime is a user-chat
service, not an engineering executor.
All command execution uses `shell=False`, fixed program allowlists, relative
paths beneath an ephemeral detached worktree, and output redaction. Direct
`git branch`, `update-ref`, `push`, `commit`, `reset`, `clean`, main,
production, tag, and baseline mutations are rejected by the worker.
## Governed write path
The worker calls the Bridge in this order:
1. `POST /v2/governed-task`;
2. `POST /v2/writer-lease`;
3. `POST /v2/commit-files`;
4. read-only authority-ref visibility verification.
CREATE_NEW is represented by the task payload and is normalized by the Bridge
lease authority. The worker never creates an authoritative ref itself.
## Lifecycle and recovery
The runtime registers with `register_or_recover`, receives its agent token
once, and stores only the encrypted systemd credential. The consumer loops over
poll, atomically claims one task, sends the first heartbeat, emits
`TASK_STARTED`, and continues heartbeats while executing.
State is atomically stored per task without agent or lease token material. If
the process stops after a governed commit but before result submission, the
next lease attempt reuses the stable governed idempotency key and submits the
stored commit evidence without issuing a second commit.
The control plane must independently verify submitted results; worker-local
success is not production or mainline authorization.
@@ -0,0 +1,3 @@
"""Generic governed Online Agent Worker runtime."""
__all__ = ["worker"]
@@ -0,0 +1,937 @@
#!/usr/bin/env python3
"""Generic, restartable Online Agent Worker.
The worker is deliberately task-agnostic. A task's signed/durable payload is
the only source of repository, branch, change, verification, and governance
parameters. There is no task id, base OID, branch name, or plan file in the
runtime configuration.
The worker performs local work only in an ephemeral detached worktree. All
authoritative ref updates go through the governed Bridge API; direct branch,
update-ref, push, main, production, tag, and baseline mutations are rejected.
"""
from __future__ import annotations
import base64
from dataclasses import dataclass
from datetime import datetime, timezone
from abc import ABC, abstractmethod
import hashlib
import json
import os
from pathlib import Path
import re
import shutil
import subprocess
import tempfile
import threading
import time
import urllib.error
import urllib.request
from typing import Any, Mapping
from urllib.parse import unquote
API_BASE = os.environ.get("AIOS_AGENT_TASK_API_URL", "http://127.0.0.1:18940").rstrip("/")
BRIDGE_BASE = os.environ.get("AIOS_GOVERNED_WRITE_API_URL", "http://127.0.0.1:18892").rstrip("/")
AGENT_ID = os.environ.get(
"AIOS_AGENT_ID", "online-engineering-worker-cn-network-mkt-generic-20260815"
)
AGENT_TYPE = os.environ.get("AIOS_AGENT_TYPE", "trusted-online-engineering-worker")
AGENT_VERSION = os.environ.get("AIOS_AGENT_VERSION", "2026-08-15.3-generic")
AUTHORITY_GIT_DIR = Path(
os.environ.get("AIOS_AUTHORITY_GIT_DIR", "/data/aios-gitops/repo.git")
)
STATE_ROOT = Path(
os.environ.get("AIOS_WORKER_STATE_ROOT", "/var/lib/aios-online-agent-worker-20260815")
)
WORKSPACE_ROOT = Path(
os.environ.get(
"AIOS_WORKER_WORKSPACE_ROOT", "/var/lib/aios-online-agent-worker-20260815/workspaces"
)
)
GIT_CONFIG_GLOBAL = STATE_ROOT / "gitconfig"
POLL_SECONDS = int(os.environ.get("AIOS_WORKER_POLL_SECONDS", "5"))
HEARTBEAT_SECONDS = int(os.environ.get("AIOS_WORKER_HEARTBEAT_SECONDS", "20"))
LEASE_SECONDS = int(os.environ.get("AIOS_WORKER_LEASE_SECONDS", "300"))
MAX_TASK_BYTES = 512 * 1024
MAX_FILE_BYTES = 2 * 1024 * 1024
OID_RE = re.compile(r"^[0-9a-f]{40}$")
SHA256_RE = re.compile(r"^[0-9a-f]{64}$")
ID_RE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._:-]{2,179}$")
BRANCH_RE = re.compile(r"^[A-Za-z0-9][A-Za-z0-9._/-]*$")
PATH_RE = re.compile(r"^[^\\\x00-\x1f\x7f]+$")
SENSITIVE_RE = re.compile(
r"(?i)(authorization\s*:\s*bearer\s+\S+|(?:password|passwd|secret|token|api[_-]?key)\s*[:=]\s*\S+)"
)
class WorkerError(RuntimeError):
def __init__(self, code: str, *, phase: str, details: Mapping[str, Any] | None = None) -> None:
super().__init__(code)
self.code = code
self.phase = phase
self.details = dict(details or {})
def now() -> str:
return datetime.now(timezone.utc).isoformat()
def digest_bytes(value: bytes) -> str:
return hashlib.sha256(value).hexdigest()
def digest_json(value: Any) -> str:
return digest_bytes(
json.dumps(value, ensure_ascii=False, sort_keys=True, separators=(",", ":")).encode("utf-8")
)
def sanitize(value: Any) -> Any:
if isinstance(value, Mapping):
cleaned: dict[str, Any] = {}
for key, inner in value.items():
lowered = str(key).lower()
if any(marker in lowered for marker in ("token", "secret", "credential", "password", "authorization")):
cleaned[str(key)] = "[REDACTED]"
else:
cleaned[str(key)] = sanitize(inner)
return cleaned
if isinstance(value, list):
return [sanitize(item) for item in value]
if isinstance(value, tuple):
return [sanitize(item) for item in value]
if isinstance(value, str):
return SENSITIVE_RE.sub("[REDACTED]", value)
return value
def excerpt(value: str, limit: int = 800) -> str:
value = str(sanitize(value))
return value if len(value) <= limit else value[: limit // 2] + "...[TRUNCATED]..." + value[-limit // 2 :]
def audit(event_type: str, **fields: Any) -> None:
STATE_ROOT.mkdir(mode=0o750, parents=True, exist_ok=True)
path = STATE_ROOT / "events.jsonl"
record = {
"event_type": event_type,
"created_at": now(),
"agent_id": AGENT_ID,
"credential_material_exposed": False,
"lease_material_exposed": False,
**sanitize(fields),
}
with path.open("a", encoding="utf-8") as handle:
handle.write(json.dumps(record, ensure_ascii=False, sort_keys=True) + "\n")
os.chmod(path, 0o640)
print(json.dumps(record, ensure_ascii=False, sort_keys=True), flush=True)
def atomic_json(path: Path, value: Mapping[str, Any], mode: int = 0o640) -> None:
path.parent.mkdir(mode=0o750, parents=True, exist_ok=True)
if path.parent.is_symlink() or path.is_symlink():
raise WorkerError("STATE_PATH_SYMLINK_REJECTED", phase="state")
fd, temporary = tempfile.mkstemp(prefix=f".{path.name}.", dir=str(path.parent))
try:
os.fchmod(fd, mode)
with os.fdopen(fd, "w", encoding="utf-8") as handle:
json.dump(sanitize(dict(value)), handle, ensure_ascii=False, sort_keys=True)
handle.write("\n")
handle.flush()
os.fsync(handle.fileno())
os.replace(temporary, path)
os.chmod(path, mode)
finally:
if os.path.exists(temporary):
os.unlink(temporary)
def safe_relative_path(value: Any, *, field: str = "path") -> str:
if not isinstance(value, str) or not value or len(value) > 240:
raise WorkerError(f"{field.upper()}_INVALID", phase="validation")
if unquote(value) != value or not PATH_RE.fullmatch(value):
raise WorkerError(f"{field.upper()}_INVALID", phase="validation")
if value.startswith(("/", "~")) or re.match(r"^[A-Za-z]:", value):
raise WorkerError(f"{field.upper()}_ABSOLUTE", phase="validation")
parts = value.replace("\\", "/").split("/")
if any(part in {"", ".", ".."} for part in parts):
raise WorkerError(f"{field.upper()}_TRAVERSAL", phase="validation")
return "/".join(parts)
def resolve_under(root: Path, relative: str, *, write: bool = False) -> Path:
relative = safe_relative_path(relative)
root = root.resolve()
candidate = (root / relative).resolve(strict=False)
if candidate != root and root not in candidate.parents:
raise WorkerError("WORKSPACE_ESCAPE", phase="validation")
current = root
for part in Path(relative).parts:
current /= part
if current.exists() and current.is_symlink():
raise WorkerError("WORKSPACE_SYMLINK_REJECTED", phase="validation")
if write and candidate.exists() and candidate.is_symlink():
raise WorkerError("WRITE_SYMLINK_REJECTED", phase="validation")
return candidate
def validate_oid(value: Any, field: str = "oid") -> str:
if not isinstance(value, str) or not OID_RE.fullmatch(value):
raise WorkerError(f"{field.upper()}_INVALID", phase="validation")
return value
def validate_sha(value: Any, field: str = "sha256") -> str:
if not isinstance(value, str) or not SHA256_RE.fullmatch(value):
raise WorkerError(f"{field.upper()}_INVALID", phase="validation")
return value
def validate_branch(value: Any) -> str:
if not isinstance(value, str) or not 3 <= len(value) <= 180:
raise WorkerError("TARGET_BRANCH_INVALID", phase="validation")
if value in {"main", "master", "HEAD"} or value.startswith(
("main/", "master/", "refs/", "tags/", "tag/", "release/", "releases/", "baseline/", "production/")
):
raise WorkerError("PROTECTED_BRANCH_REJECTED", phase="validation")
if (
value.startswith("/")
or value.endswith(("/", ".", ".lock"))
or ".." in value
or "//" in value
or "@{" in value
or any(ch in value for ch in " ~^:?*[\\")
or any(part in {"", ".", ".."} for part in value.split("/"))
or not BRANCH_RE.fullmatch(value)
):
raise WorkerError("TARGET_BRANCH_INVALID", phase="validation")
return value
def validate_program(argv: Any, *, internal_git: bool = False) -> list[str]:
if not isinstance(argv, list) or not argv or not all(isinstance(item, str) and item for item in argv):
raise WorkerError("STRUCTURED_ARGV_REQUIRED", phase="validation")
command = argv[0]
if command not in {"/usr/bin/git", "/usr/bin/python3"}:
raise WorkerError("PROGRAM_NOT_ALLOWLISTED", phase="validation", details={"program": Path(command).name})
if command.endswith("python3"):
if any(item in {"-c", "-", "--command"} for item in argv):
raise WorkerError("PYTHON_INLINE_CODE_REJECTED", phase="validation")
if "-m" in argv:
index = argv.index("-m")
if index + 1 >= len(argv) or argv[index + 1] not in {"pytest", "py_compile"}:
raise WorkerError("PYTHON_MODULE_NOT_ALLOWLISTED", phase="validation")
if command.endswith("git") and not internal_git:
rejected = {"push", "update-ref", "receive-pack", "send-pack", "reset", "clean", "branch", "commit", "merge", "rebase"}
if any(item in rejected for item in argv[1:]):
raise WorkerError("GIT_MUTATION_NOT_ALLOWLISTED", phase="validation")
return list(argv)
def run_command(
label: str,
argv: list[str],
*,
cwd: Path,
timeout: int,
internal_git: bool = False,
allowed_returncodes: tuple[int, ...] = (0,),
) -> dict[str, Any]:
validate_program(argv, internal_git=internal_git)
if not 1 <= timeout <= 3600:
raise WorkerError("COMMAND_TIMEOUT_INVALID", phase=label)
started = time.monotonic()
try:
completed = subprocess.run(
argv,
cwd=str(cwd),
text=True,
capture_output=True,
check=False,
shell=False,
timeout=timeout,
env={"PATH": "/usr/bin:/bin", "LC_ALL": "C.UTF-8", "GIT_CONFIG_GLOBAL": str(GIT_CONFIG_GLOBAL)},
)
except subprocess.TimeoutExpired as exc:
raise WorkerError("COMMAND_TIMEOUT", phase=label, details={"timeout_seconds": timeout}) from exc
result = {
"label": label,
"argv": [Path(item).name if index == 0 else item for index, item in enumerate(argv)],
"exit_code": completed.returncode,
"stdout_sha256": digest_bytes(completed.stdout.encode("utf-8", "replace")),
"stderr_sha256": digest_bytes(completed.stderr.encode("utf-8", "replace")),
"stdout_excerpt": excerpt(completed.stdout),
"stderr_excerpt": excerpt(completed.stderr),
"duration_ms": int((time.monotonic() - started) * 1000),
}
audit("COMMAND_RESULT", **result)
if completed.returncode not in allowed_returncodes:
raise WorkerError("COMMAND_FAILED", phase=label, details={"exit_code": completed.returncode, "stderr": excerpt(completed.stderr, 400)})
return result
@dataclass(frozen=True)
class ExecutionEnvelope:
task_id: str
repository: str
base_oid: str
target_branch: str
adapter: str
change_set: tuple[dict[str, Any], ...]
verification: tuple[dict[str, Any], ...]
governance: dict[str, Any]
workspace: dict[str, Any]
digest: str
@classmethod
def from_task(cls, task: Mapping[str, Any]) -> "ExecutionEnvelope":
task_id = str(task.get("task_id") or "")
if not ID_RE.fullmatch(task_id):
raise WorkerError("TASK_ID_INVALID", phase="validation")
metadata = task.get("metadata")
if not isinstance(metadata, Mapping):
raise WorkerError("TASK_METADATA_REQUIRED", phase="validation")
execution = metadata.get("execution")
if not isinstance(execution, Mapping):
raise WorkerError("DYNAMIC_EXECUTION_PAYLOAD_REQUIRED", phase="validation")
if execution.get("schema") != "aios.online-agent.execution.v2":
raise WorkerError("EXECUTION_SCHEMA_INVALID", phase="validation")
repository = execution.get("repository")
if not isinstance(repository, str) or not repository:
raise WorkerError("REPOSITORY_INVALID", phase="validation")
base_oid = validate_oid(execution.get("base_oid"), "base_oid")
target_branch = validate_branch(execution.get("target_branch"))
workspace = execution.get("workspace")
if not isinstance(workspace, Mapping):
raise WorkerError("WORKSPACE_POLICY_REQUIRED", phase="validation")
workspace_value = dict(workspace)
if workspace_value.get("authority_git_dir") != str(AUTHORITY_GIT_DIR):
raise WorkerError("AUTHORITY_ROOT_INVALID", phase="validation")
if workspace_value.get("read_only_repository") is not True:
raise WorkerError("REPOSITORY_WRITE_POLICY_INVALID", phase="validation")
if workspace_value.get("cleanup_policy") not in {"always", "cleanup_on_success"}:
raise WorkerError("WORKSPACE_CLEANUP_POLICY_INVALID", phase="validation")
adapter_value = execution.get("adapter")
if isinstance(adapter_value, Mapping):
adapter = str(adapter_value.get("name") or "")
else:
adapter = str(adapter_value or "")
if adapter not in {"deterministic", "codex", "hermes"}:
raise WorkerError("RUNTIME_ADAPTER_UNSUPPORTED", phase="validation")
change_set = execution.get("change_set")
if not isinstance(change_set, list) or not change_set or len(change_set) > 128:
raise WorkerError("CHANGE_SET_INVALID", phase="validation")
normalized_changes: list[dict[str, Any]] = []
for item in change_set:
if not isinstance(item, Mapping):
raise WorkerError("CHANGE_OPERATION_INVALID", phase="validation")
operation = str(item.get("operation") or "")
relative = safe_relative_path(item.get("path"))
normalized = dict(item)
normalized["operation"] = operation
normalized["path"] = relative
if operation == "write_file":
content = item.get("content")
if not isinstance(content, str) or len(content.encode("utf-8")) > MAX_FILE_BYTES:
raise WorkerError("WRITE_CONTENT_INVALID", phase="validation")
elif operation == "replace_text":
if not isinstance(item.get("old"), str) or not item.get("old") or not isinstance(item.get("new"), str):
raise WorkerError("REPLACE_TEXT_INVALID", phase="validation")
if len(str(item["old"]).encode("utf-8")) > MAX_FILE_BYTES:
raise WorkerError("REPLACE_TEXT_INVALID", phase="validation")
elif operation == "delete_file":
pass
else:
raise WorkerError("CHANGE_OPERATION_UNSUPPORTED", phase="validation", details={"operation": operation})
expected = item.get("expected_sha256")
if expected is not None:
validate_sha(expected, "expected_sha256")
normalized_changes.append(normalized)
verification = execution.get("verification") or []
if not isinstance(verification, list) or len(verification) > 128:
raise WorkerError("VERIFICATION_INVALID", phase="validation")
normalized_verification: list[dict[str, Any]] = []
for item in verification:
if not isinstance(item, Mapping):
raise WorkerError("VERIFICATION_COMMAND_INVALID", phase="validation")
label = str(item.get("label") or "verification")
argv = validate_program(item.get("argv"))
cwd_value = item.get("cwd")
cwd = "." if cwd_value in {None, "", "."} else safe_relative_path(cwd_value, field="cwd")
timeout = item.get("timeout_seconds", 120)
if not isinstance(timeout, int) or isinstance(timeout, bool) or not 1 <= timeout <= 3600:
raise WorkerError("VERIFICATION_TIMEOUT_INVALID", phase="validation")
normalized_verification.append({"label": label, "argv": argv, "cwd": cwd, "timeout_seconds": timeout})
governance = execution.get("governance")
if not isinstance(governance, Mapping):
raise WorkerError("GOVERNANCE_POLICY_REQUIRED", phase="validation")
governance_value = dict(governance)
governed_task_id = str(governance_value.get("task_id") or f"{task_id}-writer")
if not ID_RE.fullmatch(governed_task_id):
raise WorkerError("GOVERNED_TASK_ID_INVALID", phase="validation")
idempotency_key = str(governance_value.get("idempotency_key") or f"{task_id}-governed-v1")
commit_key = str(governance_value.get("commit_idempotency_key") or f"{task_id}-commit-v1")
for field, value in (("idempotency_key", idempotency_key), ("commit_idempotency_key", commit_key)):
if not re.fullmatch(r"[A-Za-z0-9][A-Za-z0-9._:-]{7,159}", value):
raise WorkerError(f"{field.upper()}_INVALID", phase="validation")
governance_value.update(
{
"task_id": governed_task_id,
"idempotency_key": idempotency_key,
"commit_idempotency_key": commit_key,
"message": str(governance_value.get("message") or f"chore(gitops): execute task {task_id}"),
}
)
for field in ("main_write", "force_push", "production_write", "tag_write", "baseline_write", "protected_ref_bypass"):
if governance_value.get(field, False) is not False:
raise WorkerError("GOVERNANCE_POLICY_REJECTED", phase="validation", details={"field": field})
allowed_files = governance_value.get("allowed_files")
if allowed_files is not None:
if not isinstance(allowed_files, list):
raise WorkerError("ALLOWED_FILES_INVALID", phase="validation")
allowed = {safe_relative_path(item, field="allowed_file") for item in allowed_files}
if not {item["path"] for item in normalized_changes}.issubset(allowed):
raise WorkerError("CHANGE_OUTSIDE_ALLOWED_FILES", phase="validation")
governance_value["allowed_files"] = sorted(allowed)
return cls(
task_id=task_id,
repository=repository,
base_oid=base_oid,
target_branch=target_branch,
adapter=adapter,
change_set=tuple(normalized_changes),
verification=tuple(normalized_verification),
governance=governance_value,
workspace=workspace_value,
digest=digest_json(execution),
)
class RuntimeAdapter(ABC):
name: str
@abstractmethod
def execute(self, envelope: ExecutionEnvelope, worktree: Path) -> dict[str, Any]:
raise NotImplementedError
class DeterministicExecutor(RuntimeAdapter):
name = "deterministic"
@staticmethod
def _write_atomic(path: Path, content: str) -> None:
path.parent.mkdir(mode=0o750, parents=True, exist_ok=True)
fd, temporary = tempfile.mkstemp(prefix=f".{path.name}.", dir=str(path.parent))
try:
with os.fdopen(fd, "w", encoding="utf-8") as handle:
handle.write(content)
handle.flush()
os.fsync(handle.fileno())
os.replace(temporary, path)
finally:
if os.path.exists(temporary):
os.unlink(temporary)
def execute(self, envelope: ExecutionEnvelope, worktree: Path) -> dict[str, Any]:
changed: set[str] = set()
for change in envelope.change_set:
relative = str(change["path"])
path = resolve_under(worktree, relative, write=True)
expected = change.get("expected_sha256")
current = path.read_bytes() if path.is_file() else None
if expected is not None and (current is None or digest_bytes(current) != expected):
raise WorkerError("WORKTREE_SOURCE_DRIFT", phase="implementing", details={"path": relative})
operation = change["operation"]
if operation == "write_file":
self._write_atomic(path, str(change["content"]))
elif operation == "replace_text":
if current is None:
raise WorkerError("REPLACE_SOURCE_MISSING", phase="implementing", details={"path": relative})
original = current.decode("utf-8")
old = str(change["old"])
count = original.count(old)
expected_count = int(change.get("count", 1))
if count != expected_count:
raise WorkerError("REPLACE_TEXT_MATCH_COUNT_INVALID", phase="implementing", details={"path": relative, "count": count})
self._write_atomic(path, original.replace(old, str(change["new"])))
elif operation == "delete_file":
if path.exists():
path.unlink()
changed.add(relative)
audit("REPOSITORY_CHANGE_APPLIED", operation=operation, path=relative, sha256=digest_bytes(path.read_bytes()) if path.exists() else None)
commands: list[dict[str, Any]] = []
if changed:
commands.append(run_command("git_intent_to_add", ["/usr/bin/git", "-C", str(worktree), "add", "-N", "--", *sorted(changed)], cwd=worktree, timeout=120, internal_git=True))
commands.append(run_command("git_diff_check", ["/usr/bin/git", "-C", str(worktree), "diff", "--check"], cwd=worktree, timeout=120))
commands.append(run_command("git_diff_stat", ["/usr/bin/git", "-C", str(worktree), "diff", "--no-ext-diff", "--stat"], cwd=worktree, timeout=120))
tests: list[dict[str, Any]] = []
for item in envelope.verification:
cwd = worktree if item["cwd"] == "." else resolve_under(worktree, item["cwd"])
tests.append(run_command(item["label"], item["argv"], cwd=cwd, timeout=item["timeout_seconds"]))
return {
"changed_files": sorted(changed),
"commands": commands,
"tests": tests,
"adapter": self.name,
}
class CodexExecutor(RuntimeAdapter):
name = "codex"
def execute(self, envelope: ExecutionEnvelope, worktree: Path) -> dict[str, Any]:
raise WorkerError(
"CODEX_ADAPTER_REQUIRES_EXPLICIT_POLICY",
phase="adapter",
details={"reason": "Codex is not allowed to mutate a governed worktree without an explicit file/command policy"},
)
class HermesExecutor(RuntimeAdapter):
name = "hermes"
def execute(self, envelope: ExecutionEnvelope, worktree: Path) -> dict[str, Any]:
raise WorkerError(
"HERMES_ENGINEERING_ADAPTER_UNAVAILABLE",
phase="adapter",
details={"reason": "The live Hermes service is a user-chat runtime, not an engineering executor"},
)
def adapter_for(name: str) -> RuntimeAdapter:
return {"deterministic": DeterministicExecutor, "codex": CodexExecutor, "hermes": HermesExecutor}[name]()
class HttpClient:
@staticmethod
def call(base: str, method: str, route: str, token: str | None, body: Mapping[str, Any] | None = None, timeout: int = 30) -> tuple[int, dict[str, Any]]:
payload = None if body is None else json.dumps(body, ensure_ascii=False, sort_keys=True, separators=(",", ":")).encode("utf-8")
headers = {"Accept": "application/json"}
if token:
headers["Authorization"] = "Bearer " + token
if payload is not None:
headers["Content-Type"] = "application/json"
request = urllib.request.Request(base + route, data=payload, headers=headers, method=method)
try:
with urllib.request.urlopen(request, timeout=timeout) as response:
value = json.loads(response.read().decode("utf-8"))
return response.status, value if isinstance(value, dict) else {}
except urllib.error.HTTPError as exc:
raw = exc.read(8192).decode("utf-8", "replace")
try:
value = json.loads(raw)
except json.JSONDecodeError:
value = {"error": "HTTP_ERROR"}
return exc.code, value if isinstance(value, dict) else {"error": "HTTP_ERROR"}
except (urllib.error.URLError, TimeoutError):
return 599, {"error": "SERVICE_UNAVAILABLE"}
class AgentTaskClient:
def __init__(self, token: str) -> None:
self.token = token
def poll(self) -> list[dict[str, Any]]:
status, response = HttpClient.call(API_BASE, "POST", "/v1/tasks/poll", self.token, {"limit": 100})
if status != 200:
raise WorkerError("TASK_POLL_FAILED", phase="poll", details={"api_status": status, "error": response.get("error")})
tasks = response.get("tasks")
return [dict(item) for item in tasks if isinstance(item, Mapping)] if isinstance(tasks, list) else []
def claim(self, task_id: str) -> tuple[int, dict[str, Any]]:
return HttpClient.call(API_BASE, "POST", f"/v1/tasks/{task_id}/claim", self.token, {"lease_seconds": LEASE_SECONDS})
def heartbeat(self, task_id: str, lease_token: str, stage: str) -> tuple[int, dict[str, Any]]:
return HttpClient.call(
API_BASE,
"POST",
f"/v1/tasks/{task_id}/heartbeat",
self.token,
{"lease_token": lease_token, "lease_seconds": LEASE_SECONDS, "progress": {"stage": stage, "worker_version": AGENT_VERSION}},
)
def result(self, task_id: str, lease_token: str, idempotency_key: str, result: Mapping[str, Any]) -> tuple[int, dict[str, Any]]:
return HttpClient.call(
API_BASE,
"POST",
f"/v1/tasks/{task_id}/result",
self.token,
{"lease_token": lease_token, "idempotency_key": idempotency_key, "result": dict(result)},
)
class BridgeClient:
def call(self, route: str, body: Mapping[str, Any]) -> tuple[int, dict[str, Any]]:
return HttpClient.call(BRIDGE_BASE, "POST", route, None, body, timeout=60)
class LeaseHeartbeat:
def __init__(self, client: AgentTaskClient, task_id: str, lease_token: str) -> None:
self.client = client
self.task_id = task_id
self.lease_token = lease_token
self.stage = "claimed"
self.failed: str | None = None
self.stop_event = threading.Event()
self.lock = threading.Lock()
self.thread = threading.Thread(target=self._run, name="online-agent-heartbeat", daemon=True)
def first(self) -> None:
status, response = self.client.heartbeat(self.task_id, self.lease_token, self.stage)
if not 200 <= status < 300:
self.failed = str(response.get("error") or status)
raise WorkerError("TASK_HEARTBEAT_FAILED", phase="heartbeat", details={"api_status": status, "error": self.failed})
audit("TASK_HEARTBEAT", task_id=self.task_id, stage=self.stage, api_status=status)
def start(self) -> None:
self.thread.start()
def set_stage(self, stage: str) -> None:
with self.lock:
self.stage = stage
def _run(self) -> None:
while not self.stop_event.wait(HEARTBEAT_SECONDS):
with self.lock:
stage = self.stage
try:
status, response = self.client.heartbeat(self.task_id, self.lease_token, stage)
if not 200 <= status < 300:
self.failed = str(response.get("error") or status)
audit("TASK_HEARTBEAT_FAILED", task_id=self.task_id, stage=stage, api_status=status, error=self.failed)
else:
audit("TASK_HEARTBEAT", task_id=self.task_id, stage=stage, api_status=status)
except Exception as exc: # pragma: no cover - defensive runtime guard
self.failed = type(exc).__name__
audit("TASK_HEARTBEAT_FAILED", task_id=self.task_id, stage=stage, error=self.failed)
def ensure(self) -> None:
if self.failed:
raise WorkerError("TASK_LEASE_LOST", phase="heartbeat", details={"error": self.failed})
def close(self) -> None:
self.stop_event.set()
if self.thread.ident is not None:
self.thread.join(timeout=2)
def load_agent_token() -> str:
directory = Path(os.environ.get("CREDENTIALS_DIRECTORY", ""))
path = directory / "agent_token"
if not path.is_file() or path.is_symlink():
raise WorkerError("AGENT_RUNTIME_CREDENTIAL_UNAVAILABLE", phase="startup")
value = path.read_text(encoding="utf-8").strip()
if len(value) < 24:
raise WorkerError("AGENT_RUNTIME_CREDENTIAL_INVALID", phase="startup")
return value
def ensure_roots() -> None:
if AUTHORITY_GIT_DIR.is_symlink() or STATE_ROOT.is_symlink() or WORKSPACE_ROOT.is_symlink():
raise WorkerError("ROOT_SYMLINK_REJECTED", phase="startup")
if not AUTHORITY_GIT_DIR.is_dir():
raise WorkerError("AUTHORITY_GIT_DIR_UNAVAILABLE", phase="startup")
STATE_ROOT.mkdir(mode=0o750, parents=True, exist_ok=True)
WORKSPACE_ROOT.mkdir(mode=0o750, parents=True, exist_ok=True)
if GIT_CONFIG_GLOBAL.is_symlink():
raise WorkerError("GIT_CONFIG_SYMLINK_REJECTED", phase="startup")
expected = f"[safe]\n\tdirectory = {AUTHORITY_GIT_DIR}\n"
if GIT_CONFIG_GLOBAL.exists() and GIT_CONFIG_GLOBAL.read_text(encoding="utf-8") != expected:
raise WorkerError("GIT_CONFIG_BINDING_CONFLICT", phase="startup")
if not GIT_CONFIG_GLOBAL.exists():
GIT_CONFIG_GLOBAL.write_text(expected, encoding="utf-8")
os.chmod(GIT_CONFIG_GLOBAL, 0o640)
def task_state_path(task_id: str) -> Path:
if not ID_RE.fullmatch(task_id):
raise WorkerError("TASK_ID_INVALID", phase="state")
return STATE_ROOT / "tasks" / f"{task_id}.json"
def load_task_state(task_id: str, execution_digest: str) -> dict[str, Any] | None:
path = task_state_path(task_id)
if not path.is_file() or path.is_symlink():
return None
try:
value = json.loads(path.read_text(encoding="utf-8"))
except Exception:
raise WorkerError("TASK_STATE_INVALID", phase="state")
if not isinstance(value, dict):
raise WorkerError("TASK_STATE_INVALID", phase="state")
if value.get("execution_digest") not in {None, execution_digest}:
raise WorkerError("TASK_PAYLOAD_DRIFT", phase="recovery")
return value
def save_task_state(task_id: str, value: Mapping[str, Any]) -> None:
atomic_json(task_state_path(task_id), value)
def new_run_root() -> Path:
ensure_roots()
root = Path(tempfile.mkdtemp(prefix="task-", dir=str(WORKSPACE_ROOT)))
os.chmod(root, 0o750)
return root
def cleanup_run_root(path: Path) -> None:
root = WORKSPACE_ROOT.resolve()
candidate = path.resolve(strict=False)
if candidate == root or root not in candidate.parents:
raise WorkerError("CLEANUP_SCOPE_INVALID", phase="cleanup")
if candidate.exists() or candidate.is_symlink():
shutil.rmtree(candidate)
def prepare_worktree(run_root: Path, base_oid: str) -> Path:
clone = run_root / "repository"
worktree = run_root / "worktree"
run_command("workspace_clone", ["/usr/bin/git", "clone", "--no-local", "--no-hardlinks", str(AUTHORITY_GIT_DIR), str(clone)], cwd=run_root, timeout=300, internal_git=True)
run_command("workspace_base_probe", ["/usr/bin/git", "-C", str(clone), "rev-parse", "--verify", base_oid + "^{commit}"], cwd=run_root, timeout=60)
run_command("workspace_worktree_create", ["/usr/bin/git", "-C", str(clone), "worktree", "add", "--detach", str(worktree), base_oid], cwd=run_root, timeout=300, internal_git=True)
return worktree
def governed_commit(envelope: ExecutionEnvelope, worktree: Path, bridge: BridgeClient, heartbeat: LeaseHeartbeat) -> dict[str, Any]:
governance = envelope.governance
create_body = {
"task_id": governance["task_id"],
"repository": envelope.repository,
"base_oid": envelope.base_oid,
"target_branch": envelope.target_branch,
"risk": str(governance.get("risk") or "engineering"),
"description": str(governance.get("description") or f"Governed execution for {envelope.task_id}"),
"idempotency_key": governance["idempotency_key"],
}
status, response = bridge.call("/v2/governed-task", create_body)
if not 200 <= status < 300:
raise WorkerError("GOVERNED_TASK_CREATE_FAILED", phase="committing", details={"api_status": status, "error": response.get("error")})
generation = int(response.get("generation") or 0)
expected_branch_oid = str(response.get("expected_branch_oid") or envelope.base_oid)
branch_mode = str(response.get("branch_mode") or "CREATE_NEW")
lease_body = {
"schema": "aios.task-writer-lease.v2",
"task_id": governance["task_id"],
"repository": envelope.repository,
"branch_mode": branch_mode,
"base_oid": envelope.base_oid,
"expected_branch_oid": expected_branch_oid,
"target_branch": envelope.target_branch,
"generation": generation,
"lease_seconds": int(governance.get("lease_seconds") or 900),
}
status, lease_response = bridge.call("/v2/writer-lease", lease_body)
if not 200 <= status < 300:
raise WorkerError("WRITER_LEASE_ACQUIRE_FAILED", phase="committing", details={"api_status": status, "error": lease_response.get("error")})
heartbeat.ensure()
files: dict[str, str] = {}
paths = governance.get("allowed_files") or [item["path"] for item in envelope.change_set]
for relative in sorted({safe_relative_path(item, field="allowed_file") for item in paths}):
path = resolve_under(worktree, relative)
if not path.is_file():
raise WorkerError("COMMIT_FILE_MISSING", phase="committing", details={"path": relative})
files[relative] = base64.b64encode(path.read_bytes()).decode("ascii")
commit_body = {
"task_id": governance["task_id"],
"repository": envelope.repository,
"expected_branch_oid": expected_branch_oid,
"target_branch": envelope.target_branch,
"generation": generation,
"idempotency_key": governance["commit_idempotency_key"],
"files": files,
"message": governance["message"],
}
status, commit_response = bridge.call("/v2/commit-files", commit_body)
if not 200 <= status < 300:
raise WorkerError("GOVERNED_COMMIT_FAILED", phase="committing", details={"api_status": status, "error": commit_response.get("error")})
commit_oid = commit_response.get("commit_oid")
if not isinstance(commit_oid, str) or not OID_RE.fullmatch(commit_oid):
raise WorkerError("GOVERNED_COMMIT_OID_INVALID", phase="committing")
visible_oid: str | None = None
for _ in range(5):
probe = subprocess.run(
["/usr/bin/git", "--git-dir", str(AUTHORITY_GIT_DIR), "rev-parse", "--verify", f"refs/heads/{envelope.target_branch}"],
text=True,
capture_output=True,
check=False,
env={"PATH": "/usr/bin:/bin", "LC_ALL": "C.UTF-8", "GIT_CONFIG_GLOBAL": str(GIT_CONFIG_GLOBAL)},
timeout=30,
)
if probe.returncode == 0:
visible_oid = probe.stdout.strip()
if visible_oid == commit_oid:
break
time.sleep(1)
if visible_oid != commit_oid:
raise WorkerError("GIT_VISIBILITY_MISMATCH", phase="verification", details={"commit_oid": commit_oid, "visible_oid": visible_oid})
return {
"governed_task_id": governance["task_id"],
"branch_mode": branch_mode,
"generation": generation,
"target_branch": envelope.target_branch,
"commit_oid": commit_oid,
"visible_oid": visible_oid,
"commit_idempotency_key": governance["commit_idempotency_key"],
"api_commit_visible": commit_response.get("commit_visible"),
"direct_main_write": False,
"force_push": False,
"production_write": False,
"protected_ref_bypass": False,
}
def build_result(task_id: str, envelope: ExecutionEnvelope, execution: Mapping[str, Any], governed: Mapping[str, Any]) -> dict[str, Any]:
return {
"status": "SUCCEEDED",
"verification_state": "VERIFIED_BY_WORKER_PENDING_SERVER_REVIEW",
"task_id": task_id,
"execution_backend": "generic-online-agent-worker-v2",
"adapter": envelope.adapter,
"execution_digest": envelope.digest,
"repository": envelope.repository,
"base_oid": envelope.base_oid,
"target_branch": envelope.target_branch,
"changed_files": execution.get("changed_files", []),
"commands": [{k: item.get(k) for k in ("label", "exit_code", "stdout_sha256", "stderr_sha256", "duration_ms")} for item in execution.get("commands", [])],
"tests": [{k: item.get(k) for k in ("label", "exit_code", "stdout_sha256", "stderr_sha256", "duration_ms")} for item in execution.get("tests", [])],
"governed": dict(governed),
"workspace_cleanup": "PASS",
"secret_material_exposed": False,
"direct_main_write": False,
"force_push": False,
"production_write": False,
"protected_ref_bypass": False,
}
class WorkerSupervisor:
def __init__(self, token: str) -> None:
self.agent = AgentTaskClient(token)
self.bridge = BridgeClient()
def process_claim(self, task: Mapping[str, Any], claim: Mapping[str, Any]) -> None:
task_id = str(claim.get("task_id") or task.get("task_id") or "")
lease_token = claim.get("lease_token")
if not isinstance(lease_token, str) or len(lease_token) < 24:
raise WorkerError("LEASE_TOKEN_NOT_RETURNED", phase="claim")
audit("TASK_CLAIMED", task_id=task_id, attempt=claim.get("attempt"), revision=claim.get("revision"))
heartbeat = LeaseHeartbeat(self.agent, task_id, lease_token)
try:
heartbeat.first()
heartbeat.set_stage("started")
heartbeat.start()
audit("TASK_STARTED", task_id=task_id, attempt=claim.get("attempt"))
envelope = ExecutionEnvelope.from_task(task)
state = load_task_state(task_id, envelope.digest) or {"task_id": task_id, "execution_digest": envelope.digest, "phase": "claimed"}
governed = state.get("governed") if isinstance(state.get("governed"), Mapping) else None
if isinstance(governed, Mapping) and isinstance(governed.get("commit_oid"), str) and OID_RE.fullmatch(governed["commit_oid"]):
result = state.get("result")
if not isinstance(result, Mapping):
result = {"status": "SUCCEEDED", "verification_state": "VERIFIED_BY_WORKER_PENDING_SERVER_REVIEW", "task_id": task_id, "execution_digest": envelope.digest, "governed": dict(governed), "recovered_without_duplicate_commit": True, "direct_main_write": False, "production_write": False, "force_push": False, "protected_ref_bypass": False}
audit("TASK_RECOVERED_COMMIT", task_id=task_id, commit_oid=governed["commit_oid"], duplicate_commit=False)
else:
run_root = new_run_root()
try:
worktree = prepare_worktree(run_root, envelope.base_oid)
state.update({"phase": "workspace_prepared", "execution_digest": envelope.digest, "attempt": claim.get("attempt")})
save_task_state(task_id, state)
adapter = adapter_for(envelope.adapter)
execution = adapter.execute(envelope, worktree)
state.update({"phase": "verification_passed", "execution": execution})
save_task_state(task_id, state)
heartbeat.ensure()
heartbeat.set_stage("committing")
governed = governed_commit(envelope, worktree, self.bridge, heartbeat)
result = build_result(task_id, envelope, execution, governed)
state.update({"phase": "commit_done", "governed": governed, "result": result})
save_task_state(task_id, state)
finally:
cleanup_run_root(run_root)
audit("WORKSPACE_CLEANED", task_id=task_id, cleanup="PASS")
heartbeat.ensure()
heartbeat.set_stage("result")
result_key = str(result.get("governed", {}).get("commit_idempotency_key") or f"{task_id}-result-{envelope.digest[:16]}")
status, response = self.agent.result(task_id, lease_token, result_key, result)
if not 200 <= status < 300:
raise WorkerError("TASK_RESULT_SUBMIT_FAILED", phase="result", details={"api_status": status, "error": response.get("error")})
audit("TASK_RESULT_SUBMITTED", task_id=task_id, api_status=status, result_idempotency_key=result_key, commit_oid=result.get("governed", {}).get("commit_oid"))
except WorkerError as exc:
audit("TASK_FAILED", task_id=task_id, error=exc.code, phase=exc.phase, details=exc.details)
failure_key = f"{task_id}-attempt-{claim.get('attempt', 0)}-failure"
failure = {
"status": "FAILED",
"verification_state": "UNVERIFIED",
"task_id": task_id,
"execution_backend": "generic-online-agent-worker-v2",
"error": exc.code,
"phase": exc.phase,
"details": exc.details,
"direct_main_write": False,
"force_push": False,
"production_write": False,
"protected_ref_bypass": False,
"secret_material_exposed": False,
}
try:
status, response = self.agent.result(task_id, lease_token, failure_key, failure)
audit("TASK_FAILURE_RESULT_SUBMITTED", task_id=task_id, api_status=status, error=response.get("error"))
except Exception as submit_exc: # pragma: no cover - defensive runtime guard
audit("TASK_FAILURE_RESULT_SUBMIT_FAILED", task_id=task_id, error=type(submit_exc).__name__)
finally:
heartbeat.close()
def once(self) -> int:
try:
tasks = self.agent.poll()
audit("TASK_POLL", visible_count=len(tasks))
except WorkerError as exc:
audit("TASK_POLL_FAILED", error=exc.code, phase=exc.phase, details=exc.details)
return 10
dynamic_tasks: list[dict[str, Any]] = []
for task in tasks:
metadata = task.get("metadata")
execution = metadata.get("execution") if isinstance(metadata, Mapping) else None
if not isinstance(execution, Mapping) or execution.get("schema") != "aios.online-agent.execution.v2":
audit("TASK_SKIPPED_NON_DYNAMIC", task_id=task.get("task_id"), reason="DYNAMIC_EXECUTION_PAYLOAD_REQUIRED")
continue
dynamic_tasks.append(task)
for task in dynamic_tasks:
task_id = str(task.get("task_id") or "")
status, claim = self.agent.claim(task_id)
if status == 409:
audit("TASK_CLAIM_SKIPPED", task_id=task_id, api_status=status, error=claim.get("error"))
continue
if status != 200:
audit("TASK_CLAIM_FAILED", task_id=task_id, api_status=status, error=claim.get("error"))
continue
self.process_claim(task, claim)
return 0
return 0
def run_forever(self) -> int:
ensure_roots()
audit("WORKER_STARTED", task_agnostic=True, api_base=API_BASE, bridge_base=BRIDGE_BASE, adapter_names=["deterministic", "codex", "hermes"], runtime_only_credentials=True)
while True:
self.once()
time.sleep(POLL_SECONDS)
def main() -> int:
try:
token = load_agent_token()
supervisor = WorkerSupervisor(token)
if os.environ.get("AIOS_WORKER_ONCE") == "1":
ensure_roots()
return supervisor.once()
return supervisor.run_forever()
except WorkerError as exc:
audit("WORKER_START_FAILED", error=exc.code, phase=exc.phase, details=exc.details)
return 20
if __name__ == "__main__":
raise SystemExit(main())
@@ -0,0 +1,157 @@
#!/usr/bin/env python3
"""Register or recover the generic Online Agent using runtime-only secrets."""
from __future__ import annotations
import json
import os
from pathlib import Path
import subprocess
import tempfile
import urllib.error
import urllib.request
from datetime import datetime, timezone
from typing import Any
BASE = os.environ.get("AIOS_AGENT_TASK_API_URL", "http://127.0.0.1:18940").rstrip("/")
AGENT_ID = os.environ.get("AIOS_AGENT_ID", "online-engineering-worker-cn-network-mkt-generic-20260815")
AGENT_TYPE = os.environ.get("AIOS_AGENT_TYPE", "trusted-online-engineering-worker")
AGENT_VERSION = os.environ.get("AIOS_AGENT_VERSION", "2026-08-15.3-generic")
PROJECTION_PROVIDER = "BridgeV2RuntimeCredentialProjection"
STATE_DIR = Path(os.environ.get("AIOS_WORKER_REGISTRATION_STATE", "/var/lib/aios-online-agent-worker-20260815/registration"))
CREDENTIAL_PATH = Path(os.environ.get("AIOS_WORKER_ENCRYPTED_CREDENTIAL", "/etc/credstore.encrypted/aios-online-agent-worker-20260815/agent_token"))
IDENTITY_PATH = STATE_DIR / "identity.json"
AUDIT_PATH = STATE_DIR / "events.jsonl"
CAPABILITIES = [
"online_agent_task_api", "task.poll", "task.claim", "task.heartbeat", "task.result",
"engineering.consumer", "engineering.execute", "workspace.worktree", "git.read", "git.diff",
"git.test", "governed.write.branch", "governed.write.file",
]
def now() -> str:
return datetime.now(timezone.utc).isoformat()
def audit(event_type: str, **fields: Any) -> None:
STATE_DIR.mkdir(mode=0o750, parents=True, exist_ok=True)
record = {"event_type": event_type, "created_at": now(), "agent_id": AGENT_ID, "credential_material_exposed": False, **fields}
with AUDIT_PATH.open("a", encoding="utf-8") as handle:
handle.write(json.dumps(record, ensure_ascii=False, sort_keys=True) + "\n")
os.chmod(AUDIT_PATH, 0o640)
print(json.dumps(record, ensure_ascii=False, sort_keys=True), flush=True)
def api_call(method: str, route: str, token: str, body: dict[str, Any] | None = None) -> tuple[int, dict[str, Any]]:
payload = None if body is None else json.dumps(body, ensure_ascii=False, sort_keys=True, separators=(",", ":")).encode("utf-8")
headers = {"Authorization": "Bearer " + token, "Accept": "application/json"}
if payload is not None:
headers["Content-Type"] = "application/json"
request = urllib.request.Request(BASE + route, data=payload, headers=headers, method=method)
try:
with urllib.request.urlopen(request, timeout=15) as response:
value = json.loads(response.read().decode("utf-8"))
return response.status, value if isinstance(value, dict) else {}
except urllib.error.HTTPError as exc:
try:
value = json.loads(exc.read(2048).decode("utf-8"))
except Exception:
value = {"error": "HTTP_ERROR"}
return exc.code, value if isinstance(value, dict) else {"error": "HTTP_ERROR"}
except (urllib.error.URLError, TimeoutError):
return 599, {"error": "AGENT_TASK_API_UNAVAILABLE"}
def read_identity() -> dict[str, Any] | None:
if not IDENTITY_PATH.is_file() or IDENTITY_PATH.is_symlink():
return None
try:
value = json.loads(IDENTITY_PATH.read_text(encoding="utf-8"))
except Exception:
return None
return value if isinstance(value, dict) else None
def identity_matches(value: dict[str, Any] | None) -> bool:
return bool(value and value.get("agent_id") == AGENT_ID and value.get("agent_type") == AGENT_TYPE and value.get("agent_version") == AGENT_VERSION and value.get("credential_projection_provider") == PROJECTION_PROVIDER and value.get("task_scope") == "dynamic")
def encrypted_credential_is_valid() -> bool:
if not CREDENTIAL_PATH.is_file() or CREDENTIAL_PATH.is_symlink():
return False
result = subprocess.run(["systemd-creds", "decrypt", "--with-key=host", str(CREDENTIAL_PATH), "/dev/null"], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, check=False, timeout=15)
return result.returncode == 0
def write_identity() -> None:
value = {"agent_id": AGENT_ID, "agent_type": AGENT_TYPE, "agent_version": AGENT_VERSION, "task_scope": "dynamic", "credential_projection_provider": PROJECTION_PROVIDER, "credential_delivery": "systemd-LoadCredentialEncrypted", "task_api_url": BASE, "task_api_schema": "aios.agent-task.v1", "execution_backend": "generic-online-agent-worker-v2", "runtime_only_secret_delivery": True, "production_write": False, "main_write": False, "force_push": False, "tag_write": False, "baseline_write": False, "writer_lease": True, "created_at": now()}
STATE_DIR.mkdir(mode=0o750, parents=True, exist_ok=True)
fd, temporary = tempfile.mkstemp(prefix="identity.", dir=str(STATE_DIR))
try:
with os.fdopen(fd, "w", encoding="utf-8") as handle:
json.dump(value, handle, ensure_ascii=False, sort_keys=True)
handle.write("\n")
os.chmod(temporary, 0o640)
os.replace(temporary, IDENTITY_PATH)
finally:
if os.path.exists(temporary):
os.unlink(temporary)
def encrypt_agent_token(agent_token: str) -> None:
CREDENTIAL_PATH.parent.mkdir(mode=0o700, parents=True, exist_ok=True)
fd, temporary = tempfile.mkstemp(prefix=".online-agent-token.", dir=str(CREDENTIAL_PATH.parent))
os.close(fd)
os.unlink(temporary)
try:
result = subprocess.run(["systemd-creds", "encrypt", "--with-key=host", "--name=agent_token", "/dev/stdin", temporary], input=(agent_token + "\n").encode("utf-8"), stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL, check=False, timeout=20)
if result.returncode != 0:
raise RuntimeError("AGENT_CREDENTIAL_ENCRYPT_FAILED")
os.chmod(temporary, 0o600)
os.replace(temporary, CREDENTIAL_PATH)
os.chmod(CREDENTIAL_PATH, 0o600)
finally:
if os.path.exists(temporary):
os.unlink(temporary)
def register_or_recover() -> str:
STATE_DIR.mkdir(mode=0o750, parents=True, exist_ok=True)
existing = read_identity()
if CREDENTIAL_PATH.exists():
if not identity_matches(existing):
raise RuntimeError("AGENT_IDENTITY_BINDING_CONFLICT")
if not encrypted_credential_is_valid():
raise RuntimeError("AGENT_ENCRYPTED_CREDENTIAL_INVALID")
audit("AGENT_RECOVERED", credential_delivery="systemd-LoadCredentialEncrypted", credential_projection_provider=PROJECTION_PROVIDER, register_or_recover=True)
return "RECOVERED"
if existing is not None:
raise RuntimeError("AGENT_CREDENTIAL_MISSING_AFTER_PRIOR_REGISTRATION")
credential_directory = Path(os.environ.get("CREDENTIALS_DIRECTORY", ""))
control_path = credential_directory / "bridge_control_token"
if not control_path.is_file():
raise RuntimeError("BRIDGE_RUNTIME_CONTROL_CREDENTIAL_UNAVAILABLE")
control_token = control_path.read_text(encoding="utf-8").strip()
if len(control_token) < 32:
raise RuntimeError("BRIDGE_RUNTIME_CONTROL_CREDENTIAL_INVALID")
status, response = api_call("POST", "/v1/agents/register", control_token, {"agent_id": AGENT_ID, "agent_type": AGENT_TYPE, "agent_version": AGENT_VERSION, "capabilities": CAPABILITIES, "metadata": {"task_scope": "dynamic", "credential_projection_provider": PROJECTION_PROVIDER, "runtime_only_secret_delivery": True, "task_api_schema": "aios.agent-task.v1", "execution_backend": "generic-online-agent-worker-v2", "production_write": False, "main_write": False, "force_push": False, "tag_write": False, "baseline_write": False, "writer_lease": True}})
if status == 409 and response.get("error") == "AGENT_ALREADY_EXISTS":
raise RuntimeError("AGENT_ALREADY_EXISTS_RECOVERY_CREDENTIAL_UNAVAILABLE")
if status != 201:
raise RuntimeError("AGENT_REGISTER_FAILED_" + str(response.get("error") or status))
agent_token = response.get("agent_token")
if not isinstance(agent_token, str) or len(agent_token) < 24:
raise RuntimeError("AGENT_TOKEN_NOT_RETURNED")
encrypt_agent_token(agent_token)
write_identity()
audit("AGENT_REGISTERED", credential_delivery="systemd-LoadCredentialEncrypted", credential_projection_provider=PROJECTION_PROVIDER, register_or_recover=True)
return "REGISTERED"
if __name__ == "__main__":
try:
raise SystemExit(0 if register_or_recover() in {"REGISTERED", "RECOVERED"} else 20)
except Exception as exc:
audit("AGENT_REGISTRATION_FAILED", error=type(exc).__name__)
raise SystemExit(20)
@@ -0,0 +1,79 @@
from __future__ import annotations
import os
from pathlib import Path
from aios_online_agent_worker.worker import ExecutionEnvelope, WorkerError, sanitize # noqa: E402
BASE = "a" * 40
def payload(task_id: str, branch: str, *, base_oid: str = BASE) -> dict[str, object]:
return {
"task_id": task_id,
"metadata": {
"execution": {
"schema": "aios.online-agent.execution.v2",
"repository": "aios/aios",
"base_oid": base_oid,
"target_branch": branch,
"adapter": "deterministic",
"workspace": {
"authority_git_dir": str(Path(os.environ.get("AIOS_AUTHORITY_GIT_DIR", "/data/aios-gitops/repo.git"))),
"read_only_repository": True,
"cleanup_policy": "always",
},
"change_set": [
{
"operation": "write_file",
"path": "tests/generated_marker.py",
"content": "def test_marker():\n assert True\n",
"expected_sha256": None,
}
],
"verification": [
{"label": "compile", "argv": ["/usr/bin/python3", "-m", "py_compile", "tests/generated_marker.py"], "cwd": ".", "timeout_seconds": 60}
],
"governance": {
"task_id": f"{task_id}-writer",
"idempotency_key": f"{task_id}-governed-v1",
"commit_idempotency_key": f"{task_id}-commit-v1",
"allowed_files": ["tests/generated_marker.py"],
"main_write": False,
"force_push": False,
"production_write": False,
"tag_write": False,
"baseline_write": False,
"protected_ref_bypass": False,
},
}
},
}
def test_dynamic_payload_is_not_bound_to_task_or_base() -> None:
first = ExecutionEnvelope.from_task(payload("task-alpha-20260815", "agent/alpha-20260815"))
second = ExecutionEnvelope.from_task(payload("task-beta-20260815", "agent/beta-20260815", base_oid="b" * 40))
assert first.task_id != second.task_id
assert first.base_oid != second.base_oid
assert first.target_branch != second.target_branch
def test_protected_branch_is_rejected() -> None:
try:
ExecutionEnvelope.from_task(payload("task-main-20260815", "main"))
except WorkerError as exc:
assert exc.code == "PROTECTED_BRANCH_REJECTED"
else:
raise AssertionError("protected branch accepted")
def test_task_payload_does_not_persist_secret_values() -> None:
value = sanitize({"agent_token": "agt-secret", "nested": {"authorization": "Bearer secret"}})
assert value == {"agent_token": "[REDACTED]", "nested": {"authorization": "[REDACTED]"}}
def test_execution_digest_is_stable() -> None:
value = payload("task-digest-20260815", "agent/digest-20260815")
assert ExecutionEnvelope.from_task(value).digest == ExecutionEnvelope.from_task(value).digest