Add trigger-api skill for Management API SDK

[oborchers]

Proposal

Add a new trigger-api skill covering the Trigger.dev Management API SDK — the full set of methods for managing runs, tasks, schedules, environment variables, queues, waitpoints, batches, deployments, and TRQL queries from backend code.

Gap

The existing 6 skills cover task definitions (trigger-tasks), configuration (trigger-config), AI agent patterns (trigger-agents), realtime subscriptions (trigger-realtime), setup (trigger-setup), and cost optimization (trigger-cost-savings) — but none cover the Management API SDK (runs.list, schedules.create, envvars.upload, query.execute, etc.).

This means AI agents have no skill to load when users need to:

• List, filter, or monitor runs from backend code
• Trigger tasks programmatically outside of other tasks
• Manage schedules, env vars, or queues via the SDK
• Create/complete waitpoint tokens from external systems
• Query run data with TRQL

What the skill covers

• SKILL.md (~290 lines): Setup, authentication (secret key vs PAT), raw HTTP usage, error handling, auto-pagination, quick reference table for all ~50 endpoints, common patterns
• references/runs-and-tasks.md (~420 lines): Task triggering (3 methods), runs CRUD (10 methods), run object reference
• references/schedules.md (~290 lines): DECLARATIVE vs IMPERATIVE, full CRUD, ScheduleObject reference
• references/queues-envvars-waitpoints.md (~400 lines): Queues, env vars, waitpoints, deployments (REST-only)
• references/batches-and-query.md (~260 lines): Two-phase batch API, TRQL query with verified column list

Tested against live API

All documented REST endpoints were tested against a live Trigger.dev dev environment. Corrections applied:

• runs.retrieve uses /api/v3/, not /api/v1/
• runs.cancel uses /api/v2/, not /api/v1/
• batches.create uses /api/v3/, not /api/v1/
• runs.retrieveResult removed (404 on all versions — phantom endpoint in docs)
• schedules.timezones marked SDK-only (no REST equivalent)
• REST-only endpoints marked (retrieveEvents, retrieveTrace, addTags, updateMetadata)
• PAT auth for runs.list marked as broken (fails via both raw HTTP and SDK)
• TRQL column list verified with actual query results (40 columns)

How it complements existing skills

• trigger-tasks = defining tasks and triggering from within tasks (myTask.trigger())
• trigger-api = managing resources from backend scripts/servers (runs.list(), schedules.create())

Happy to iterate on structure or content based on maintainer feedback.

[oborchers]

Keeping my pr in draft mode. The most present issue is that the list runs API returns HTML. This would simplify my setup a bit