Investigation: Campaign worker workflow labeling system

[Copilot]

Documents how campaigns apply labels to worker outputs for discovery and coordination.

Labeling System

Workers apply two labels via safe-outputs configuration:

• agentic-campaign - Generic marker preventing non-campaign workflow interference
• z_campaign_<campaign-id> - Campaign-specific identifier for discovery (lowercase, hyphenated)

safe-outputs:
  create-pull-request:
    labels: [security, automated-fix, agentic-campaign, z_campaign_security-alert-burndown]

Discovery Flow

Campaign orchestrators search GitHub using label:"z_campaign_<campaign-id>" as primary discovery mechanism. Label format enforced consistently across:

• Go: pkg/stringutil/identifiers.go:FormatCampaignLabel()
• JS: actions/setup/js/campaign_discovery.cjs
• JS: actions/setup/js/safe_output_handler_manager.cjs

Deliverable

Added docs/investigation-campaign-worker-labels.md (628 lines) covering:

• Label types, format, and application mechanisms
• Data flow architecture
• Safe outputs configuration reference
• Code locations and examples
• Best practices and common pitfalls

The z_ prefix ensures campaign labels sort last in GitHub's UI for visibility.

Original prompt

Investigate how campaigns currently set worker workflows labels (agentic-campaign, z_campaign, ...)

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.