Orchestrate a full engine cycle — pre-flight, pre-mortem, engine run, engine review, world summary.
Orchestrate a full engine cycle — pre-flight, pre-mortem, engine run, engine review, world summary.
npx skills add https://github.com/pnils08/GodWorld --skill run-cycleこのコマンドをClaude Codeにコピー＆ペーストしてスキルをインストール
Audit architecture and system docs for staleness, inaccuracy, and drift from actual codebase state. Grouped into 7 tiers for thorough coverage. Complement to /md-audit (existence-staleness — when that ships).
Prepare all inputs for city-hall voice agents. Reads tracker, approvals, world summary, engine review, coverage ratings, previous log, canon, Mara directive. Writes pending decisions per voice.
Run the civic government — voice agents make decisions, tracker updates, canon output. Mags sifts, agents decide.
Immersive scene piece — one reporter, one location, one moment. Editions report, supplementals explain, dispatches immerse.
Post-publish print pipeline — photos, PDF, Drive upload. Edition-only for now (S188 rebuild); non-edition types await T11 wiring.
Capture-only — transcript becomes canon. Articles framed off interview transcripts move downstream to /write-edition (next cycle, via /sift) or /write-supplemental (any time, sports-desk subagent dispatch against the canon transcript).
| name | run-cycle |
| description | Orchestrate a full engine cycle — pre-flight, pre-mortem, engine run, engine review, world summary. |
| version | 1.1 |
| updated | "2026-05-24T00:00:00.000Z" |
| tags | ["engine-sheet","active"] |
| effort | high |
| disable-model-invocation | true |
| argument-hint | [cycle-number] |
Runs the full engine cycle pipeline by calling individual skills in order. Each step produces output on disk that the next step reads.
Verify manual inputs are ready — sports feed entries, initiative tracker structure, coverage ratings from last edition. Citizen/business/storyline intake not wired yet (placeholder).
Gate: READY → proceed. NOT READY → stop and fix.
Scan engine code for determinism violations, dependency chain breaks, sheet header misalignment, cascade risks.
Gate: CLEAN → proceed. CRITICAL → stop and fix.
Tell Mike to run runWorldCycle() in Google Apps Script. The engine runs in Google's cloud — cannot be triggered from here. Wait for confirmation it finished.
Gate: Mike confirms engine completed.
Read world state from sheets. Identify ailments, improvements, incoherence. Produce 7-field briefs per finding. Output: output/engine_review_c{XX}.md
Gate: File exists on disk.
Read Riley_Digest (3 cycles), Sports Feed (3 cycles), civic production log (if exists), and engine review output. Produce factual world summary. Output: output/world_summary_c{XX}.md. Ingest to world-data Supermemory.
Gate: File exists on disk + ingest confirmed.
Run the mechanical baseline audit and append judgment-layer entries.
node scripts/engineCycleAudit.js {XX} --write
Writes output/production_log_c{XX}_run_cycle_gaps.md with [mechanical]-tagged entries across 5 detector classes (writeback-drift, math-anomaly, cross-cycle-debt, determinism-break, header-drift). 4 V2-runtime classes (phase-skip, cohort-collision, phase-ordering, silent-fail) appended as stubs — they need an engine-run-log ingest path that doesn't exist yet.
header-drift (S202 build) is a structural class — runs against repo state vs schemas/SCHEMA_HEADERS.md, flags writers whose field-name lookups or setValues() ranges don't match live headers. Self-test: node scripts/engineCycleAuditTest.js (synthetic fixture validating the detector would have caught the S201 Story_Seed_Deck/Story_Hook_Deck drift). Plan: docs/plans/2026-05-05-writer-header-alignment-detector.md.
After the script runs, engine-sheet appends [judgment] entries below the <!-- end mechanical pass --> footer marker. The script preserves anything below that marker on re-run, so judgment entries survive a re-audit pass.
Coder-persona voice for judgment entries: terse, mechanical, commit-message style (no narrative prose). G-EC{N+} numbering continues from the mechanical pass.
If 0 mechanical entries observed: file states "0 mechanical gaps observed" with cycle headline metrics for context (vs being absent, which would be ambiguous). Engine-sheet judgment review still recommended.
Gate: File exists on disk. ROLLOUT pointer added under Edition Post-Publish for any HIGH-severity findings: RUN-CYCLE C{XX} GAP LOG — S{N} (engine-sheet), N entries, severity.
Plan: docs/plans/2026-05-03-run-cycle-gap-log-surface.md (Phase 2 done S199; Phase 3 validation runs at next /run-cycle invocation).
These run as separate skills (may be same or different sessions):
/city-hall-prep — reads world summary + engine review + sheets → writes pending decisions per voice/city-hall — reads pending decisions → launches voice agents → applies tracker updates → production log/sift (planned) — reads world summary + engine review + city-hall log → story picks + angle briefs/write-edition — reads sift output → launches reporters → compile → publishHandoff: run-cycle produces output/world_summary_c{XX}.md and output/engine_review_c{XX}.md. City-hall-prep verifies both exist before starting.
The following scripts were part of the old inline pipeline. Not called by this skill but preserved for future use:
scripts/buildDeskPackets.js — per-desk JSON packets from cycle packetscripts/buildDeskFolders.js — per-desk workspace foldersscripts/buildInitiativePackets.js — per-initiative JSON packetsscripts/buildInitiativeWorkspaces.js — per-initiative workspace foldersscripts/buildVoiceWorkspaces.js — per-voice-agent workspace foldersscripts/buildDecisionQueue.js — pending decisions for voice agentsscripts/checkSupplementalTriggers.js — supplemental edition candidatesService account via lib/sheets.js. Spreadsheet ID from .env.