openspec-sync

name	openspec-sync
description	Update docs and context for OpenSpec changes. Use when: saving session state, refreshing docs after implementation, or syncing change status.
model	sonnet
allowed-tools	["Glob","Grep","Read","Edit","Write","Bash"]

OpenSpec Sync

Context and documentation synchronization for OpenSpec changes. Keeps CONTEXT-llm.md and docs in sync with implementation state.

Workflow: Check → Gather → Write

flowchart LR
    A["Get status via CLI"] --> B["Read current state"]
    B --> C["Gather context"]
    C --> D["Write updates"]
    D --> E["Verify idempotent"]

    classDef action fill:#C8E6C9,stroke:#388E3C,color:#000
    class A,B,C,D,E action

Critical: Use openspec status for change state, never parse files directly.

Guardrails

Autonomous (no confirmation needed):

• CONTEXT-llm.md updates
• Doc content refresh
• Status queries via CLI

Ask-first (pause and confirm):

• Creating new doc files not mentioned in proposal
• Deleting outdated docs
• Major restructuring of existing docs

CLI Delegation

Always use CLI for status:

openspec status {change-id}

Never:

• Parse proposal.md for status manually
• Infer status from file existence
• Maintain separate status tracking

The CLI is the single source of truth for change status.

Commands

sync

Full synchronization: updates CONTEXT-llm.md and refreshes docs in one command.

Input: $ARGUMENTS = change-id

Workflow:

1. Run openspec status {change-id} for current state
2. Read git state: git status --short, git log -1 --oneline
3. Read openspec/changes/{change-id}/tasks.md for progress
4. Update CONTEXT-llm.md with gathered state
5. Identify docs affected by implementation (Glob for modified files)
6. Refresh affected docs if out of sync

Output:

✅ Sync complete for {change-id}
- CONTEXT-llm.md: updated
- Docs refreshed: {count} files
- Change status: {status from CLI}

docs

Refresh documentation files only (no CONTEXT-llm.md update).

Input: $ARGUMENTS = change-id

Workflow:

1. Run openspec status {change-id} for affected files
2. Read openspec/changes/{change-id}/proposal.md → Affected Files
3. Glob for modified files since change started
4. Identify docs that reference changed files
5. Update those docs to reflect current implementation

Output:

📚 Docs refresh for {change-id}
- Scanned: {n} affected files
- Updated: {m} docs
- Skipped: {k} (already current)

context

Save session state to CONTEXT-llm.md only (no docs refresh).

Input: $ARGUMENTS = change-id

Workflow:

1. Run openspec status {change-id}
2. Run git status --short
3. Run git log -3 --oneline
4. Read openspec/changes/{change-id}/tasks.md
5. Count completed vs total tasks
6. Write CONTEXT-llm.md with template below

CONTEXT-llm.md Template:

# Context: {change-id}

## Status
- Change: {status from openspec status}
- Tasks: {done}/{total} complete
- Git: {branch}, {clean|modified}

## Recent Commits
{last 3 commits}

## Current Focus
{current task or section being worked on}

## Next Steps
1. {next unchecked task}
2. {following task}

## Reference
- Proposal: openspec/changes/{change-id}/proposal.md
- Tasks: openspec/changes/{change-id}/tasks.md
- Project: openspec/project.md

Output:

💾 Context saved for {change-id}
- File: CONTEXT-llm.md
- Tasks: {done}/{total}
- Status: {status}

Philosophy Check

Before syncing, read openspec/project.md → Execution Philosophy → mode.

In garage mode: Minimal docs, just CONTEXT-llm.md. Don't gold-plate documentation. In scale mode: Ensure all docs accurate and up-to-date. In maintenance mode: Document changes carefully before modifying existing docs.

Idempotency

Running sync multiple times with same state produces identical output:

1. Before writing, read current CONTEXT-llm.md
2. Generate new content
3. Compare: if identical, report "Already in sync"
4. Write only if different

Output when no changes:

⏭️ Already in sync for {change-id}
- CONTEXT-llm.md: current
- Docs: current
- Last sync: {timestamp from file}

Exploration Strategy

Before syncing, consult openspec/project.md → Exploration Strategy section:

1. Context sources: Read primary files (project.md, proposal.md, tasks.md)
2. Must-read files: CLAUDE.md, settings.json (project constraints)
3. Tools: Use configured codebase tools (Glob, Grep, Read)
4. Philosophy: Read Execution Philosophy section for current mode