# Grimblaz Workflow Template Parallel Execution > Parallel Execution Skill - Skill: `tomevault-io/grimblaz-workflow-template-parallel-execution` (Agent Skill, multi-file: 2 files) - Install (CLI): `npx skillmds@latest add tomevault-io/grimblaz-workflow-template-parallel-execution` - Raw SKILL.md: https://api.skillmd.com/api/skills/tomevault-io/grimblaz-workflow-template-parallel-execution/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Productivity - Author: tomevault-io (https://skillmd.com/u/tomevault-io) - Updated: 2026-09-17 - Page: https://skillmd.com/skills/tomevault-io/grimblaz-workflow-template-parallel-execution --- # Parallel Execution Skill Execution protocol for implementation steps that may run in parallel or serial mode. ## Mode Declaration For each implementation step, declare one mode before delegation: - `Execution Mode: parallel` - `Execution Mode: serial` Record the declaration in: 1. Conductor progress log header 2. Step metadata/state (`step.metadata.execution_mode`) 3. User-facing progress update 4. Step commit details if tracked in the plan ## Requirement Contract (Mandatory) Before delegation, define a shared Requirement Contract for the step: 1. Acceptance criteria slice for the step 2. Explicit invariants and edge cases 3. Non-goals for this step Both implementation and testing lanes must use this same contract. ## Protocol 1. **Choose mode** using stability/risk criteria. 2. **Create Requirement Contract** before any lane starts. 3. **Run lanes**: - `parallel`: launch Code-Smith and Test-Writer against the same contract. - `serial`: run one lane first, then the second against the same contract. 4. **Run triage** via Test-Writer after outputs are available. 5. **Classify failures** as `code defect`, `test defect`, `harness/env defect`, or `rc-divergence` with evidence. 6. **Route corrections bidirectionally** until convergence: - `code defect` → Code-Smith - `test defect` → Test-Writer - `harness/env defect` → responsible specialist/tooling path - `rc-divergence` → conditional sequential pair: Code-Smith first (fix implementation to match RC); CC re-evaluates; if resolved → advance; if not → Test-Writer with instruction: "Re-derive test assertions from the Requirement Contract, not from the corrected implementation." 7. **Enforce convergence gate** before advancing. 8. **Run RC conformance check** — CC evaluates the step's Requirement Contract AC items against delivered code after convergence. Divergences route as `rc-divergence` with the conditional sequential pair described in step 6. ## Convergence Gate (Mandatory) Do not advance phase until Test-Writer explicitly confirms: - Green tests - Valid assertions - No brittle coupling After convergence, the RC conformance check (protocol step 8) verifies that the step's Requirement Contract AC items are satisfied before step advance. ## Loop Budget - Maximum 3 correction cycles per step. - If exceeded, perform root-cause review and escalate via `#tool:vscode/askQuestions` with a recommended option. - RC conformance correction uses 1 dedicated cycle outside the main 3-cycle budget. If unresolved after 1 cycle, escalate via `#tool:vscode/askQuestions` with unresolved AC items and recommended options. ## Anti-Test-Chasing Guardrail Code-Smith must satisfy the Requirement Contract and architecture constraints, not merely optimize for currently failing assertions. If a property/test appears over-constrained or invalid, classify as potential `test defect` and route back for test review with evidence. ## Post-Issue Checkpoint (Mandatory) At issue/PR completion, record a short process checkpoint: - What slowed us down? - What failed late that should fail earlier? - What single guardrail should be added next? Track this in the completion summary to improve future cycles. ## Gotchas | Trigger | Gotcha | Fix | | ------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Test-Writer writes tests to match Code-Smith's implementation instead of the Requirement Contract | Tests pass but AC is not met; the test is wrong, not the implementation | Route back as `test defect` with evidence when assertion appears over-constrained relative to AC | | Starting implementation lanes before the Requirement Contract is agreed | Lanes diverge against different assumptions; reconciliation is expensive | Mandate Requirement Contract before any lane begins — block execution until it exists | | Advancing to the next phase on a verbal "looks good" | Missing explicit green/valid/non-brittle confirmation; state is ambiguous | Block phase advance until all 3 convergence criteria explicitly stated in writing | | More than 3 correction cycles with no resolution | Infinite loop; work stalls; effort compounds with no progress | At cycle 3, perform root-cause analysis and escalate via `#tool:vscode/askQuestions` with concrete options | | Classifying a test defect as a code defect (or vice versa) | Wrong agent receives the correction task; fix misses the actual problem | Triage before routing: `code defect`, `test defect`, `harness/env defect`, or `rc-divergence` — require evidence | | Skipping the post-issue process checkpoint after merge | Lessons from slow steps and late failures are lost | Record the 3-question checkpoint in every completion summary | | RC conformance flags divergence but CC routes as `code defect` instead of `rc-divergence` | Code-Smith fixes code without Test-Writer re-deriving from RC; tests still reflect divergent behavior | Always route RC conformance divergences as `rc-divergence` — conditional sequential pair (Code-Smith → re-evaluate → Test-Writer with "re-derive from RC" if divergence persists) | --- > Converted and distributed by [TomeVault](https://tomevault.io/claim/grimblaz) — claim your Tome and manage your conversions.