security: add TTL lease for claim branches, vulnerability scanning cron, and improvements\n\n- Implement claim TTL (2h) to prevent stale locks\n- Add global cleanup of expired claim branches\n- Add daily security_scan.py using pip-audit; schedule via OpenClaw cron\n- Monitor-prs now checks claim expiration and cleans up globally\n- Improves resilience of multi-agent coordination

This commit is contained in:
2026-03-16 07:35:33 +00:00
parent 258b320d3a
commit f11f277e71
3 changed files with 202 additions and 22 deletions

44
dev/scripts/security_scan.py Executable file
View File

@@ -0,0 +1,44 @@
#!/usr/bin/env python3
"""
Security vulnerability scanner for AITBC dependencies.
Uses pip-audit to check installed packages in the CLI virtualenv.
"""
import subprocess
import json
import sys
PIP_AUDIT = '/opt/aitbc/cli/venv/bin/pip-audit'
def run_audit():
try:
result = subprocess.run([PIP_AUDIT, '--format', 'json'],
capture_output=True, text=True, timeout=300)
if result.returncode not in (0, 1): # 1 means vulns found, 0 means clean
return f"❌ pip-audit execution failed (exit {result.returncode}):\n{result.stderr}"
data = json.loads(result.stdout) if result.stdout else {}
vulns = data.get('vulnerabilities', [])
if not vulns:
return "✅ Security scan: No known vulnerabilities in installed packages."
# Summarize by severity
sev_counts = {}
for v in vulns:
sev = v.get('severity', 'UNKNOWN')
sev_counts[sev] = sev_counts.get(sev, 0) + 1
lines = ["🚨 Security scan: Found vulnerabilities:"]
for sev, count in sorted(sev_counts.items(), key=lambda x: x[1], reverse=True):
lines.append(f"- {sev}: {count} package(s)")
# Add top 3 vulnerable packages
if vulns:
lines.append("\nTop vulnerable packages:")
for v in vulns[:3]:
pkg = v.get('package', 'unknown')
vuln_id = v.get('vulnerability_id', 'unknown')
lines.append(f"- {pkg}: {vuln_id}")
return "\n".join(lines)
except Exception as e:
return f"❌ Error during security scan: {str(e)}"
if __name__ == '__main__':
message = run_audit()
print(message)
sys.exit(0)

View File

@@ -2,11 +2,12 @@
"""
Task Claim System for AITBC agents.
Uses Git branch atomic creation as a distributed lock to prevent duplicate work.
Now with TTL/lease: claims expire after 2 hours to prevent stale locks.
"""
import os
import json
import subprocess
from datetime import datetime
from datetime import datetime, timezone
REPO_DIR = '/opt/aitbc'
STATE_FILE = '/opt/aitbc/.claim-state.json'
@@ -16,6 +17,7 @@ MY_AGENT = os.getenv('AGENT_NAME', 'aitbc1')
ISSUE_LABELS = ['security', 'bug', 'feature', 'refactor', 'task'] # priority order
BONUS_LABELS = ['good-first-task-for-agent']
AVOID_LABELS = ['needs-design', 'blocked', 'needs-reproduction']
CLAIM_TTL_SECONDS = 7200 # 2 hours lease
def query_api(path, method='GET', data=None):
url = f"{API_BASE}/{path}"
@@ -88,6 +90,24 @@ def claim_issue(issue_number):
result = subprocess.run(['git', 'push', 'origin', branch_name], capture_output=True, text=True, cwd=REPO_DIR)
return result.returncode == 0
def is_claim_stale(claim_branch):
"""Check if a claim branch is older than TTL (stale lock)."""
try:
result = subprocess.run(['git', 'ls-remote', '--heads', 'origin', claim_branch],
capture_output=True, text=True, cwd=REPO_DIR)
if result.returncode != 0 or not result.stdout.strip():
return True # branch missing, treat as stale
# Optional: could check commit timestamp via git show -s --format=%ct <sha>
# For simplicity, we'll rely on state file expiration
return False
except Exception:
return True
def cleanup_stale_claim(claim_branch):
"""Delete a stale claim branch from remote."""
subprocess.run(['git', 'push', 'origin', '--delete', claim_branch],
capture_output=True, cwd=REPO_DIR)
def assign_issue(issue_number, assignee):
data = {"assignee": assignee}
return query_api(f'repos/oib/aitbc/issues/{issue_number}/assignees', method='POST', data=data)
@@ -105,17 +125,35 @@ def create_work_branch(issue_number, title):
return branch_name
def main():
now = datetime.utcnow().isoformat() + 'Z'
print(f"[{now}] Claim task cycle starting...")
now = datetime.utcnow().replace(tzinfo=timezone.utc)
now_iso = now.isoformat()
now_ts = now.timestamp()
print(f"[{now_iso}] Claim task cycle starting...")
state = load_state()
current_claim = state.get('current_claim')
# Check if our own claim expired
if current_claim:
claimed_at = state.get('claimed_at')
expires_at = state.get('expires_at')
if expires_at and now_ts > expires_at:
print(f"Claim for issue #{current_claim} has expired (claimed at {claimed_at}). Releasing.")
# Delete the claim branch and clear state
claim_branch = state.get('claim_branch')
if claim_branch:
cleanup_stale_claim(claim_branch)
state = {}
save_state(state)
current_claim = None
if current_claim:
print(f"Already working on issue #{current_claim} (branch {state.get('work_branch')})")
# Optional: could check if that PR has been merged/closed and release claim here
return
# Optional global cleanup: delete any stale claim branches (older than TTL)
cleanup_global_stale_claims(now_ts)
issues = get_open_unassigned_issues()
if not issues:
print("No unassigned issues available.")
@@ -126,25 +164,70 @@ def main():
title = issue['title']
labels = [lbl['name'] for lbl in issue.get('labels', [])]
print(f"Attempting to claim issue #{num}: {title} (labels={labels})")
# Check if claim branch exists and is stale
claim_branch = f'claim/{num}'
if not is_claim_stale(claim_branch):
print(f"Claim failed for #{num} (active claim exists). Trying next...")
continue
# Force-delete any lingering claim branch before creating our own
cleanup_stale_claim(claim_branch)
if claim_issue(num):
assign_issue(num, MY_AGENT)
work_branch = create_work_branch(num, title)
expires_at = now_ts + CLAIM_TTL_SECONDS
state.update({
'current_claim': num,
'claim_branch': f'claim/{num}',
'claim_branch': claim_branch,
'work_branch': work_branch,
'claimed_at': datetime.utcnow().isoformat() + 'Z',
'claimed_at': now_iso,
'expires_at': expires_at,
'issue_title': title,
'labels': labels
})
save_state(state)
print(f"✅ Claimed issue #{num}. Work branch: {work_branch}")
add_comment(num, f"Agent `{MY_AGENT}` claiming this task. (automated)")
print(f"✅ Claimed issue #{num}. Work branch: {work_branch} (expires {datetime.fromtimestamp(expires_at, tz=timezone.utc).isoformat()})")
add_comment(num, f"Agent `{MY_AGENT}` claiming this task with TTL {CLAIM_TTL_SECONDS/3600}h. (automated)")
return
else:
print(f"Claim failed for #{num} (branch exists). Trying next...")
print(f"Claim failed for #{num} (push error). Trying next...")
print("Could not claim any issue; all taken or unavailable.")
def cleanup_global_stale_claims(now_ts=None):
"""Remove claim branches that appear stale (based on commit age)."""
if now_ts is None:
now_ts = datetime.utcnow().timestamp()
# List all remote claim branches
result = subprocess.run(['git', 'ls-remote', '--heads', 'origin', 'claim/*'],
capture_output=True, text=True, cwd=REPO_DIR)
if result.returncode != 0 or not result.stdout.strip():
return
lines = result.stdout.strip().split('\n')
cleaned = 0
for line in lines:
if not line.strip():
continue
parts = line.split()
if len(parts) < 2:
continue
sha, branch = parts[0], parts[1]
# Get commit timestamp
ts_result = subprocess.run(['git', 'show', '-s', '--format=%ct', sha],
capture_output=True, text=True, cwd=REPO_DIR)
if ts_result.returncode == 0 and ts_result.stdout.strip():
commit_ts = int(ts_result.stdout.strip())
age = now_ts - commit_ts
if age > CLAIM_TTL_SECONDS:
print(f"Expired claim branch: {branch} (age {age/3600:.1f}h). Deleting.")
cleanup_stale_claim(branch)
cleaned += 1
if cleaned == 0:
print(" cleanup_global_stale_claims: none")
else:
print(f" cleanup_global_stale_claims: removed {cleaned} expired branch(es)")
if __name__ == '__main__':
main()

View File

@@ -4,14 +4,14 @@ Enhanced monitor for Gitea PRs:
- Auto-request review from sibling on my PRs
- Auto-validate sibling's PRs and approve if passing checks, with stability ring awareness
- Monitor CI statuses and report failures
- Release claim branches when associated PRs merge or close
- Release claim branches when associated PRs merge, close, or EXPIRE
"""
import os
import json
import subprocess
import tempfile
import shutil
from datetime import datetime
from datetime import datetime, timezone
GITEA_TOKEN = os.getenv('GITEA_TOKEN') or 'ffce3b62d583b761238ae00839dce7718acaad85'
REPO = 'oib/aitbc'
@@ -19,6 +19,7 @@ API_BASE = os.getenv('GITEA_API_BASE', 'http://gitea.bubuit.net:3000/api/v1')
MY_AGENT = os.getenv('AGENT_NAME', 'aitbc1')
SIBLING_AGENT = 'aitbc' if MY_AGENT == 'aitbc1' else 'aitbc1'
CLAIM_STATE_FILE = '/opt/aitbc/.claim-state.json'
CLAIM_TTL_SECONDS = 7200 # Must match claim-task.py
def query_api(path, method='GET', data=None):
url = f"{API_BASE}/{path}"
@@ -74,6 +75,14 @@ def release_claim(issue_number, claim_branch):
save_claim_state(state)
print(f"✅ Released claim for issue #{issue_number} (deleted branch {claim_branch})")
def is_claim_expired(state):
"""Check if the current claim has exceeded TTL."""
expires_at = state.get('expires_at')
if not expires_at:
return False
now_ts = datetime.utcnow().timestamp()
return now_ts > expires_at
def get_open_prs():
return query_api(f'repos/{REPO}/pulls?state=open') or []
@@ -126,23 +135,30 @@ def validate_pr_branch(pr):
shutil.rmtree(tmpdir, ignore_errors=True)
def main():
now = datetime.utcnow().isoformat() + 'Z'
print(f"[{now}] Monitoring PRs and claim locks...")
now = datetime.utcnow().replace(tzinfo=timezone.utc)
now_iso = now.isoformat()
now_ts = now.timestamp()
print(f"[{now_iso}] Monitoring PRs and claim locks...")
# 0. Check claim state: if we have a current claim, see if corresponding PR merged
# 0. Check claim state: if we have a current claim, see if it expired or PR merged
state = load_claim_state()
if state.get('current_claim'):
issue_num = state['current_claim']
work_branch = state.get('work_branch')
claim_branch = state.get('claim_branch')
# Check expiration
if is_claim_expired(state):
print(f"Claim for issue #{issue_num} has expired. Releasing.")
release_claim(issue_num, claim_branch)
else:
# Check if PR merged/closed
all_prs = get_all_prs(state='all')
matched_pr = None
for pr in all_prs:
if pr['head']['ref'] == work_branch:
matched_pr = pr
break
if matched_pr:
if matched_pr['state'] == 'closed':
if matched_pr and matched_pr['state'] == 'closed':
release_claim(issue_num, claim_branch)
# 1. Process open PRs
@@ -191,10 +207,47 @@ def main():
for s in failing:
notifications.append(f"PR #{number} status check failure: {s.get('context','unknown')} - {s.get('status','unknown')}")
# 2. Global cleanup of stale claim branches (orphaned, older than TTL)
cleanup_global_expired_claims(now_ts)
if notifications:
print("\n".join(notifications))
else:
print("No new alerts.")
def cleanup_global_expired_claims(now_ts=None):
"""Delete remote claim branches that are older than TTL, even if state file is gone."""
if now_ts is None:
now_ts = datetime.utcnow().timestamp()
# List all remote claim branches
result = subprocess.run(['git', 'ls-remote', '--heads', 'origin', 'claim/*'],
capture_output=True, text=True, cwd='/opt/aitbc')
if result.returncode != 0 or not result.stdout.strip():
return
lines = result.stdout.strip().split('\n')
cleaned = 0
for line in lines:
if not line.strip():
continue
parts = line.split()
if len(parts) < 2:
continue
sha, branch = parts[0], parts[1]
# Get commit timestamp
ts_result = subprocess.run(['git', 'show', '-s', '--format=%ct', sha],
capture_output=True, text=True, cwd='/opt/aitbc')
if ts_result.returncode == 0 and ts_result.stdout.strip():
commit_ts = int(ts_result.stdout.strip())
age = now_ts - commit_ts
if age > CLAIM_TTL_SECONDS:
print(f"Expired claim branch: {branch} (age {age/3600:.1f}h). Deleting.")
subprocess.run(['git', 'push', 'origin', '--delete', branch],
capture_output=True, cwd='/opt/aitbc')
cleaned += 1
if cleaned == 0:
print(" cleanup_global_expired_claims: none")
else:
print(f" cleanup_global_expired_claims: removed {cleaned} expired branch(es)")
if __name__ == '__main__':
main()