sync-template（同步模板）

Sync Template

Synchronize a downstream project generated from blogic-template-ts with the latest upstream template state.

Use blogic-cz/blogic-template-ts as source of truth.

Prefer this local path when available:

opensrc/repos/github.com/blogic-cz/blogic-template-ts/

Reference that path as $TPL in commands.

Prerequisites

• Use gh CLI authenticated when available.
• Load debugging-with-opensrc skill when available; otherwise run equivalent opensrc / git commands directly.
• Use agent-tools (gh-tool release list/view) when available; otherwise use native gh release list/view.
• Clone template with opensrc when available; otherwise clone https://github.com/blogic-cz/blogic-template-ts into a temporary local directory and use that path as $TPL.

Phase 0: Discovery — What Changed in the Template

Step 0: Materialize template source locally

Materialize the template repo before diffing:

bun run opensrc:use blogic-cz/blogic-template-ts
ls opensrc/repos/github.com/blogic-cz/blogic-template-ts/

If opensrc is unavailable, clone with git clone and set $TPL to that clone path.

Step 1: Detect current template version in project

Detect the current template version marker:

• package.json (templateVersion, template-version, synced-from)
• .template-version
• Git history for sync commits
• CHANGELOG.md sync entries

If no baseline exists, request the baseline version from the user before continuing.

Step 2: Enumerate upstream changes since baseline

Enumerate changes with releases, commits, and PR metadata:

gh-tool release list --repo blogic-cz/blogic-template-ts
gh-tool release view <tag> --repo blogic-cz/blogic-template-ts
gh pr list --repo blogic-cz/blogic-template-ts --state merged --search "merged:>YYYY-MM-DD" --limit 50

When gh-tool is unavailable, replace gh-tool commands with gh release... commands.

Step 3: Categorize changes for adoption strategy

Classify each change into one category:

Output a summary table:

| # | Category | Change | Template Files | Effort | Priority |
|---|----------|--------|---------------|--------|----------|

Phase 1: Package Updates First (MANDATORY)

Align package versions before applying template file sync.

Step 1: Compare project and template package versions

Read package manifests in both trees and build a version-gap table.

Build a comparison table:

| Package | Project Version | Template Version | Gap |
|---------|----------------|-----------------|-----|

Step 2: Execute package updates

If packages are behind, prefer update-packages skill. If that skill is unavailable, run equivalent manual updates with the same constraints.

Apply this sequence:

1. Create branch: chore/sync-template-packages-$(date +%y%m%d-%H%M)
2. Update only already-used packages to template version or newer.
3. Commit package updates separately before template file sync commits.
4. Verify with project checks before continuing.

Apply these rules:

• Package updates MUST be committed before any template changes
• Preserve package-group coordination (tanstack, trpc, effect, drizzle)
• DO NOT add packages the project doesn't use — only update existing ones

Phase 2: Deep Scan & Diff

Compare the project against the template file by file to find all divergences.

Step 1: Identify infrastructure files to adopt

Load references/template-files.md for the categorized file list.

For each infrastructure file:

1. Check if the file exists in the project
2. If it exists, diff it against the template version
3. If it doesn't exist, flag it as "missing from project"

git diff --no-index $TPL/<file> <file>

Step 2: Build the name substitution matrix

Build substitution matrix per references/substitution-rules.md.

Step 3: Identify project-specific divergences to preserve

Flag intentional business-logic divergences and preserve them:

• apps/web-app/src/routes/ — project-specific routes
• packages/services/src/ — project-specific services
• packages/db/src/schema/ — project-specific schema (beyond template defaults)
• Environment-specific configs (Helm values, CI secrets)

Apply these rules:

• NEVER overwrite project-specific business logic with template defaults
• Template infrastructure files (lint, tooling, CI) should be adopted fully
• When a file has BOTH infrastructure changes AND project-specific content, merge selectively

Phase 3: Triage & Plan

Generate a phased implementation plan in .sisyphus/plans/.

Step 1: Create the plan file

mkdir -p .sisyphus/plans

Name: .sisyphus/plans/YYMMDDHHMM-sync-template-vX.Y.Z.md

Step 2: Apply canonical plan format

Load and apply references/plan-template.md.

Step 3: Wave organization

Standard wave pattern for template sync:

Wave 0 (Sequential — foundation):
└── Package updates (if not done in Phase 1)

Wave 1 (Parallel — infrastructure):
├── Skills & tooling updates
├── Lint & formatter config
├── CI/CD pipeline updates
├── Docker & Helm updates
└── Config file updates (tsconfig, vitest, etc.)

Wave 2 (After Wave 1 — code changes):
├── Fix lint violations from new rules
├── Apply code pattern updates
└── Update docs (AGENTS.md, README)

Wave 3 (After Wave 2 — cleanup):
└── Final cleanup + verification

Wave FINAL (After ALL — verification):
├── Plan compliance audit
├── Code quality review
└── Scope fidelity check

Step 4: Present plan to user

Present summary and wait for explicit approval before execution:

## Sync Template v<old> → v<new>

**Changes**: N commits, M PRs, K releases
**Waves**: 4 (+ verification)
**Tasks**: N total
**Estimated effort**: [Small/Medium/Large]

Approve to start execution?

Phase 4: Execution

Execute the plan wave by wave.

Prefer task delegation when subagents are available.

Fallback when subagents are unavailable: execute tasks sequentially in the main agent, keep the same acceptance criteria, and preserve planned commit grouping.

Per-task execution flow

For each task in the plan:

1. Mark TODO in_progress
2. Execute task (delegate when available; otherwise run directly)
3. Verify acceptance criteria
4. Run QA scenario
5. Mark TODO completed
6. Commit as specified in plan

Delegation prompt template

Use the delegation prompt from references/plan-template.md.

After each wave

bun run check
bun run test

Resolve failures before the next wave.

Phase 5: Verification & Stamp

Run verification recipes from references/verification-recipes.md.

Apply substitution policy from references/substitution-rules.md.

Guardrails

• NEVER overwrite project-specific business logic (routes, services, schema beyond defaults)
• NEVER leave @blogic-template/ or blogic-template-ts strings in synced project files, except approved exceptions in references/substitution-rules.md
• ALWAYS clone the template via bun run opensrc:use blogic-cz/blogic-template-ts first
• ALWAYS apply the name substitution matrix to every file copied from the template
• ALWAYS update packages FIRST before applying infrastructure changes
• ALWAYS run bun run check && bun run test after each wave
• ALWAYS preserve project-specific configurations (Sentry project, Helm namespaces, CI secrets)
• DO NOT add template-only dependencies the project doesn't need
• DO NOT apply template features the project hasn't opted into
• DO NOT use hardcoded absolute paths — always use relative opensrc/repos/github.com/blogic-cz/blogic-template-ts/

Definition of Done

• bun run check passes (0 errors)
• bun run test passes
• No @blogic-template/ or blogic-template-ts strings in synced project files, except approved exceptions in references/substitution-rules.md
• .template-version file updated to new version
• All plan tasks marked completed
• Plan file exists in .sisyphus/plans/