flow-studio-swarm

Review system calibration from run data in RUN_BASE. Use when analyzing patterns across runs for improvement opportunities, triggered by Wisdom flow or manual review.

Review new dependencies for security and maintenance risk. Use when adding or updating dependencies.

Ensure safe deployments with rollback planning. Use when preparing deployments or reviewing deploy readiness.

Plan and document deprecations with staged lifecycle, migration guide, and tooling requirements. Use when deprecating or breaking compatibility.

Route known failure patterns to standard fixes. Use when matching error signatures to detour catalog.

Classify and route errors by category (transient, permanent, retriable, fatal). Use when diagnosing failures, deciding retry vs escalate, or triggering circuit breaker.

Turn claims into evidence-bound verification. Use when validating tests/lint/security claims, gate decisions, or reviewer checklists.

Safe git operations for branch management and history. Use when working with branches, commits, or repo operations.

Refactor project guidance out of CLAUDE.md into thin always-on rules, on-demand skills, and canonical docs. Use when reducing context bloat, reorganizing rules, cleaning documentation, or deciding where new guidance belongs.

Review pull requests against intent and evidence. Use when reviewing PRs or validating changes before merge.

Execute rollbacks safely. Use when reverting deployments, commits, or configuration changes.

Resume interrupted runs from last checkpoint in RUN_BASE. Use when continuing failed or interrupted flow execution, restoring from receipts and handoffs.

Write executable runbook checklists. Use when creating operational procedures or playbooks.

Validate runbooks are executable and complete. Use when reviewing or testing runbook quality.

Handle secret exposure incidents. Use when credentials are leaked, committed to git, or detected in logs/outputs.

Rotate secrets and credentials safely. Use when rotating API keys, passwords, or handling scheduled credential updates.

Configure appropriate timeouts for operations. Use when setting up execution limits or diagnosing timeout issues.