| name | magic-run |
| description | Workflow for executing tasks from the project plan. |
| handoffs | [{"label":"Replan","workflow":"magic-task","prompt":"Revalidate the plan against current specs. Pre-Planning Stabilization auto-fixes mechanical drift; HALT with a single /magic-spec recommendation only on substantive design gaps. Per `rules/MAGIC-md §5`, this is the SOLE user-visible next step after a phase — never propose /magic-analyze or /magic-spec proactively.","condition":null}] |
Run Workflow
Triggers: "Start tasks", "Next task", "Continue", "Start parallel execution", "Launch agents", "Implement", "Apply", "Run", "Execute", "Start work".
Trigger: /magic-run [arg]. Arguments:
- (empty) — full execution across all workspaces
{workspace} — scoped execution within a workspace
"text" — directed execution (task ID, phase, or focus)
{workspace} "text" — scoped + directed execution
Examples: /magic-run, /magic-run engine, /magic-run "T-1A01", /magic-run "phase-2", /magic-run docs "validation tasks only".
Scope: code execution, task state management, and changelog generation. Task generation/orchestration is handled by magic-task. Specification authoring by magic-spec. Health auditing by magic-analyze.
- Rules Parity: always check for version mismatch between
RULES.md and TASKS.md before starting.
- Pipeline:
magic-spec → magic-task → magic-run.
- Finalization: after task/phase completion, run
node .magic/scripts/executor.js finalize --workflow=run and display output verbatim. Never auto-commit. See .magic/run.md §Finalization Protocol.
Full implementation: .magic/run.md. Read it before proceeding.
Executor: node .magic/scripts/executor.js <script> for all automation.
Anti-Hallucination Guard: do not invent ad-hoc scripts (.js, .sh, etc.) for internal engine operations. Magic SDD steps are evaluated cognitively unless an executor script is explicitly provided.