| name | sprint-runner |
| description | Sequential scenario build executor. Iterates through an approved automation portfolio, builds each scenario in Make.com one at a time with narration and per-scenario confirmation, writes logs, and produces a factory report. Called by scenario-orchestrator during the Sprint phase. |
Skill: sprint-runner
Executes an approved automation portfolio sequentially. Called by scenario-orchestrator
after the user has approved the full batch design. This skill owns the Make.com write
operations — all prior phases are read-only.
Sub-files
SPRINT-PHASES.md — Input contract, opening announcement, error handling protocol, user controls
SPRINT-BUILD.md — Per-automation build sequence and completion report format
Deterministic Classification — This Skill Owns Writes
This skill is the ONLY place in the /factory pipeline where non-deterministic calls
are made. It runs ONLY during status: "sprint" in the factory session.
Deterministic (call freely):
mcp__claude_ai_Make__scenarios_get
mcp__claude_ai_Make__scenarios_list
mcp__claude_ai_Make__connections_list
mcp__claude_ai_Make__hooks_get
Non-deterministic Level 1 (pre-execute hook gates these):
mcp__claude_ai_Make__scenarios_create
mcp__claude_ai_Make__scenarios_activate
mcp__claude_ai_Make__scenarios_update
mcp__claude_ai_Make__hooks_create
Non-deterministic Level 2 (HIGH RISK — never called automatically):
mcp__claude_ai_Make__scenarios_run ← only if user explicitly asks to test-run
Non-deterministic Level 3 — PROHIBITED in this skill entirely:
All *_delete tools
If current-session.json status is anything other than "sprint", this skill
must refuse non-deterministic calls and return control to the orchestrator.
Output Contract
Returns to orchestrator:
{
"built": [...],
"skipped": [...],
"failed": [...],
"total_ops_added": 0,
"report_path": ".make/factory/{session_id}-report.md"
}