Campaigns

Campaigns are declarative JSON definitions that describe automated tasks for agents to execute. Each campaign specifies triggers, tools, process steps, guardrails, and success metrics.

Active Campaigns

Maintenance & Monitoring

Campaign	Agent	Schedule	Budget	Description
cc-gateway-health	gateway-direct	Hourly	5K tokens	Verifies all 43 MCP tools respond correctly
cc-mcp-regression	gateway-direct	On push	5K tokens	Regression tests for tool schemas and behavior
oc-gateway-smoketest	openclaw	Manual	10K tokens	Quick 3-tool validation of gateway health
oc-dep-audit	openclaw	Weekly Mon 2am	100K tokens	Cross-repo dependency version divergence analysis
xa-audit-completeness	cross-agent	Dependent	20K tokens	Validates audit coverage across all tool calls

Security

Campaign	Agent	Schedule	Budget	Description
hs-cred-exposure	hexstrike	Weekly Sun 1am	80K tokens	Credential exposure scan across 10 repos
hs-cve-monitor	hexstrike	Daily 3am	60K tokens	CVE monitoring against dependency manifests
hs-dep-vuln	hexstrike	Dependent on dep-audit	80K tokens	Deep vulnerability analysis of flagged dependencies

Analysis & Reporting

Campaign	Agent	Schedule	Budget	Description
oc-weekly-digest	openclaw	Weekly Mon 6am	30K tokens	Aggregated weekly ecosystem summary
oc-issue-triage	openclaw	Daily 8am	40K tokens	Classify and triage untriaged issues
oc-prompt-audit	openclaw	Monthly 1st	50K tokens	Self-improvement: analyze campaign quality
oc-docs-freshness	openclaw	Weekly	40K tokens	Documentation staleness across 11 repos
oc-coverage-gaps	openclaw	Weekly	40K tokens	Test coverage gap analysis
oc-license-scan	openclaw	Weekly	30K tokens	License compliance for OSS dependencies

Campaign Definition Schema

Campaign definitions are JSON files in test/campaigns/. Key fields:

{
  "id": "campaign-id",
  "name": "Human-Readable Name",
  "agent": "openclaw|hexstrike|gateway-direct",
  "trigger": {
    "schedule": "0 6 * * 1",
    "event": "manual|push|pull_request|dependent",
    "dependsOn": ["other-campaign-id"],
    "pathFilters": ["gateway/*.go"]
  },
  "tools": ["juggler_campaign_status", "github_fetch"],
  "process": ["Step 1: ...", "Step 2: ..."],
  "guardrails": {
    "maxDuration": "15m",
    "readOnly": true,
    "killSwitch": "remotejuggler/campaigns/global-kill",
    "aiApiBudget": { "maxTokens": 30000 }
  },
  "feedback": {
    "createIssues": true,
    "closeResolvedIssues": true
  },
  "metrics": {
    "kpis": ["repos_scanned", "findings_count"]
  }
}

See test/campaigns/schema.json for the full JSON Schema definition.

Execution Flow

1. Campaign Runner evaluates triggers every 60 seconds
2. Due campaigns are dispatched to the appropriate agent
3. Agent executes process steps using MCP tools through rj-gateway
4. Results are stored in Setec (latest + timestamped history)
5. Feedback handler creates/closes GitHub issues for findings
6. Publisher posts sanitized results to GitHub Discussions
7. README agent status table is updated via repository_dispatch