| name | agent-label-routing |
| description | Deterministic multi-agent task assignment via GitHub labels — classify, label, and generate queue views from `agent:` labels instead of manual queue files |
| version | 1.0.0 |
| category | coordination |
| type | skill |
| trigger | manual |
| auto_execute | false |
| capabilities | ["agent_classification","batch_labeling","queue_generation","overnight_planning"] |
| tools | ["terminal","gh"] |
| scripts | ["refresh-agent-work-queue.sh"] |
Agent Label Routing
Deterministic agent task assignment using GitHub agent: labels. Labels ARE the queue — separate markdown files are just filtered views.
Label Schema
| Label | Color | Role |
|---|
agent:gemini | F5A623 | Research, prep, large-doc ingestion, standards mapping |
agent:claude | DA552F | Heavy coding, architecture, orchestration, complex TDD |
agent:codex | 3182CE | Bounded implementation, test writing, review, refactoring |
agent:any | 8E54E9 | No strong preference — whichever agent has capacity |
Workflow
Step 1: Create Labels (if missing)
gh label create "agent:gemini" --color "F5A623" --description "Research, prep, large-doc ingestion"
gh label create "agent:claude" --color "DA552F" --description "Heavy coding, architecture, orchestration"
gh label create "agent:codex" --color "3182CE" --description "Bounded implementation, tests, review"
gh label create "agent:any" --color "8E54E9" --description "No strong preference"
Step 2: Classify Issues
Use rule-based scoring with keyword matching + category boosts:
keywords: literature, review, research, catalog, triage, summarize,
standards gap, standards mapping, migrate literature, acquire, index,
scrape, dedup, job market
category boost: cat:document-intelligence (+3), cat:data-pipeline + dark-intelligence (+2)
keywords: architecture, orchestration, governance, credit utilization,
work queue, model switching, dispatch, integration, concept selection,
capex/opex, facility sizing, production profile, floating platform,
stability, gyradius, trim ballast, epic
category boost: cat:ai-orchestration (+), cat:engineering + dark-intelligence (+)
keywords: test coverage, bug fix, solver queue, extract metadata,
batch-at-stop, config-protection hook, convert agent to skill,
deduplicate skill, provider config
category boost: cat:engineering-calculations + domain:code-promotion (+)
