super/box
1
0
Fork 0
Code Issues 1 Pull Requests Actions Packages Projects Releases Wiki Activity

fix(dispatch): eliminate noisy SSH signatures from routine job prompts and activate real tool triggers

Browse Source
This commit is contained in:
operator
2026-10-05 04:04:41 +00:00
parent d09e1c4872
commit dc1b4f8ce6
3 changed files with 46 additions and 43 deletions
Download Patch File Download Diff File Expand all files Collapse all files
bin/job-dispatch.py
+38 -37
View File
@@ -222,44 +222,45 @@ def send_dm(agent, target, message, dry_run=False, followup_tags=None,
if HAS_RATE_LIMITER: if HAS_RATE_LIMITER:
rate_limit_wait(agent) rate_limit_wait(agent)
# Cryptographic attestation: sign message with local SSH key # Cryptographic attestation: only sign if explicitly requested by job config
# to avoid blowing up agent context windows with massive base64 SSH signature blocks.
signed_payload = None signed_payload = None
dm_sign_sh = NETVM_ROOT / "bin" / "dm-sign.sh" if os.environ.get("JOB_REQUIRE_SIGNATURE") == "1":
priv_key = Path(os.path.expanduser("~/.ssh/id_ed25519")) dm_sign_sh = NETVM_ROOT / "bin" / "dm-sign.sh"
if dm_sign_sh.exists() and priv_key.exists(): priv_key = Path(os.path.expanduser("~/.ssh/id_ed25519"))
try: if dm_sign_sh.exists() and priv_key.exists():
sign_res = subprocess.run( try:
[str(dm_sign_sh), "--from", "super", "--key", str(priv_key), message], sign_res = subprocess.run(
capture_output=True, text=True, timeout=10 [str(dm_sign_sh), "--from", "super", "--key", str(priv_key), message],
) capture_output=True, text=True, timeout=10
if sign_res.returncode == 0 and "-----BEGIN SSH SIGNATURE-----" in sign_res.stdout: )
signed_payload = sign_res.stdout.strip() if sign_res.returncode == 0 and "-----BEGIN SSH SIGNATURE-----" in sign_res.stdout:
# Extract message id and register proof to crypt.muse-dev.online signed_payload = sign_res.stdout.strip()
id_m = re.search(r"\[id:([a-f0-9]+)\]", signed_payload) id_m = re.search(r"\[id:([a-f0-9]+)\]", signed_payload)
proof_id = id_m.group(1) if id_m else None proof_id = id_m.group(1) if id_m else None
if proof_id: if proof_id:
proof_data = { proof_data = {
"id": proof_id, "id": proof_id,
"signer": "super", "signer": "super",
"target_agent": agent, "target_agent": agent,
"target_conversation": target, "target_conversation": target,
"raw_payload": signed_payload, "raw_payload": signed_payload,
"ts": datetime.now(timezone.utc).isoformat() "ts": datetime.now(timezone.utc).isoformat()
} }
try: try:
import urllib.request import urllib.request
req = urllib.request.Request( req = urllib.request.Request(
"https://crypt.muse-dev.online/proofs", "https://crypt.muse-dev.online/proofs",
data=json.dumps(proof_data).encode("utf-8"), data=json.dumps(proof_data).encode("utf-8"),
headers={"Content-Type": "application/json", "User-Agent": "job-dispatch/1.0"}, headers={"Content-Type": "application/json", "User-Agent": "job-dispatch/1.0"},
method="POST" method="POST"
) )
with urllib.request.urlopen(req, timeout=3) as resp: with urllib.request.urlopen(req, timeout=3) as resp:
pass pass
except Exception as pe: except Exception as pe:
sys.stderr.write(f"warning: proof registration to crypt.muse-dev.online failed: {pe}\n") sys.stderr.write(f"warning: proof registration to crypt.muse-dev.online failed: {pe}\n")
except Exception as se: except Exception as se:
sys.stderr.write(f"warning: dm signing failed: {se}\n") sys.stderr.write(f"warning: dm signing failed: {se}\n")
payload_to_send = signed_payload or message payload_to_send = signed_payload or message
jobs/box-http-health.json
+3 -2
View File
@@ -11,11 +11,12 @@
}, },
"name": "box-http-health", "name": "box-http-health",
"on_failure": "alert", "on_failure": "alert",
"prompt_template": "Box HTTP health check.\nJob ID: {job_id}\nTime: {datetime}\n\nRun the Box HTTP checks on the VM:\n ssh dev-operator-646@34.139.37.135 \"/srv/box/bin/box-health-check.sh http\"\n(or run /srv/box/bin/box-health-check.sh http via any VM access you have)\n\nAlso verify the new live Box Console endpoints:\n1. https://box.muse-dev.online/ (Front-Door Console loads, assets box.js and box.css return 200)\n2. https://box.muse-dev.online/api/box/fleet (Returns 200 with live fleet array)\n3. https://box.muse-dev.online/api/box/dm/log (Returns 200 with recent DM log entries)\n\nExpected: all OK. If any FAIL, investigate immediately (check board.service,\nCaddy, recent deploys) and include the failure lines in your reply.\n\nReply with OK or FAIL <one-line summary>.", "prompt_template": "Box HTTP endpoint health check.\nJob ID: {job_id}\nTime: {datetime}\n\nRun the fleet health check:\n[TOOL health.check {}]\n\nCheck recent scheduled runs:\n[TOOL cron.runs {}]\n\nIf all endpoints and fleet nodes return green, reply with [RESULT {job_id}] OK.\nOtherwise reply with [RESULT {job_id}] FAIL <summary>.",
"schedule": "*/15 * * * *", "schedule": "*/15 * * * *",
"sidechat": { "sidechat": {
"create": true, "create": true,
"name_template": "box-http-health-{datetime}" "name_template": "box-http-health-{datetime}",
"reuse_key": "box-http-health"
}, },
"timeout": 600 "timeout": 600
} }
jobs/box-service-health.json
+4 -3
View File
@@ -1,7 +1,7 @@
{ {
"agent": "646", "agent": "646",
"chain_next": null, "chain_next": null,
"description": "Box service + data health — every 15 minutes (offset 7m)", "description": "Box service + data health \u2014 every 15 minutes (offset 7m)",
"followup": { "followup": {
"escalate": "opm", "escalate": "opm",
"expect_reply": true, "expect_reply": true,
@@ -11,11 +11,12 @@
}, },
"name": "box-service-health", "name": "box-service-health",
"on_failure": "alert", "on_failure": "alert",
"prompt_template": "Box service health check.\nJob ID: {job_id}\nTime: {datetime}\n\nRun the Box service and data checks on the VM:\n ssh dev-operator-646@34.139.37.135 \"/srv/box/bin/box-health-check.sh services\"\n ssh dev-operator-646@34.139.37.135 \"/srv/box/bin/box-health-check.sh data\"\n(or run /srv/box/bin/box-health-check.sh via any VM access you have)\n\nExpected: board.service active, caddy active, sweeper timer firing,\n/srv/box writable. If any FAIL, investigate and include failure lines.\n\nReply with [RESULT {job_id}] OK or [RESULT {job_id}] FAIL <one-line summary>.", "prompt_template": "Box service and data health check.\nJob ID: {job_id}\nTime: {datetime}\n\nExecute the service check on the VM:\n[TOOL service.status {\"unit\": \"board.service\"}]\n\nCheck caddy and harvester timer status:\n[TOOL service.status {\"unit\": \"response-harvester.timer\"}]\n\nIf healthy, end with [RESULT {job_id}] OK.\nIf any service fails, reply with [RESULT {job_id}] FAIL <summary>.",
"schedule": "7,22,37,52 * * * *", "schedule": "7,22,37,52 * * * *",
"sidechat": { "sidechat": {
"create": true, "create": true,
"name_template": "box-service-health-{datetime}" "name_template": "box-service-health-{datetime}",
"reuse_key": "box-service-health"
}, },
"timeout": 600 "timeout": 600
} }