sync
Scannednpx machina-cli add skill anombyte93/atlas-session-lifecycle/sync --openclawSync
Zero questions. Edit files. Print summary. Done.
Rules
- No questions — never ask the user anything. Just save.
- No subagents — run entirely in the main thread.
- No MCP tools — direct file reads and edits only (faster).
- No narration — don't announce steps. Just do them.
- Touch every file — even if only adding a timestamp entry.
- Speed over perfection — a rough sync beats no sync.
Execute
Use the current time formatted as HH:MM DD/MM/YY for the checkpoint header.
1. Edit session-context/CLAUDE-activeContext.md
Append a new section at the end:
## [SYNC] {time}
**Accomplished:**
- {bullet list of what was done this session}
**In progress:**
- {what's currently being worked on}
**Next steps:**
- {what comes next}
**Blockers:**
- {any blockers, or "None"}
2. Edit session-context/CLAUDE-decisions.md
Append new decisions made this session. If none, append:
## {time}
No new decisions this session.
3. Edit session-context/CLAUDE-patterns.md
Append new patterns discovered. If none, append:
## {time}
No new patterns this session.
4. Edit session-context/CLAUDE-troubleshooting.md
Append new issues and solutions. If none, append:
## {time}
No new issues this session.
5. Update MEMORY.md
Update the auto-memory file at the project memory path with any new learnings from this session. If nothing new, skip this file (memory should stay concise).
6. Print summary
Print exactly one block:
Synced. {N} files updated.
- Active context: {1-line summary of what was captured}
- Decisions: {count new or "no new"}
- Patterns: {count new or "no new"}
- Troubleshooting: {count new or "no new"}
- Memory: {updated or "no changes"}
/sync --full (Capability Inventory)
Extended sync that includes capability inventory generation.
Execute
- Run standard sync flow (Steps 1-5 above).
- Call
session_capability_inventory(project_dir, force_refresh=True)— bypasses cache and regenerates inventory even if git unchanged. - If
needs_generation == True: the MCP tool will generate the inventory. - Read
CLAUDE-capability-inventory.mdif it exists.
Print summary
Print the standard sync block above, plus:
Capability Inventory: {status}
- MCP Tools: {count}
- Test Coverage: {percentage}
- Critical Untested: {count}
- Inventory: {path}
Source
git clone https://github.com/anombyte93/atlas-session-lifecycle/blob/main/skills/sync/SKILL.mdView on GitHub Overview
Sync is a fast, no-questions save-point that updates all session-context files and MEMORY.md to reflect current progress. It runs without prompts, uses direct file edits in the main thread, and is triggered by /sync or /sync --full for capability inventory.
How This Skill Works
The tool appends a time-stamped [SYNC] section to CLAUDE-activeContext.md with Accomplished, In progress, Next steps, and Blockers. It then appends the time stamp to CLAUDE-decisions.md, CLAUDE-patterns.md, and CLAUDE-troubleshooting.md, and updates MEMORY.md. Finally, it prints a one-block summary. With /sync --full, it also regenerates a capability inventory via session_capability_inventory and includes a Capability Inventory section in the summary.
When to Use It
- User explicitly invokes /sync to save progress and sync context.
- You need an up-to-date capability inventory via /sync --full.
- You require a zero-question, instant persistence workflow.
- You want to ensure every relevant file is touched, even if no new data is present.
- You need a concise, auditable memory update after a session.
Quick Start
- Step 1: Issue the /sync command to save progress and context.
- Step 2: Read the printed summary block showing updates and memory changes.
- Step 3: If needed, run /sync --full to regenerate and review the Capability Inventory.
Best Practices
- Always use /sync to create a checkpoint of current progress.
- Do not prompt the user; keep the operation in the main thread.
- Append accurate time headers and avoid unnecessary edits.
- Review the summary block and confirm MEMORY.md reflects new learnings.
- Use /sync --full when inventory regeneration is required or beneficial.
Example Use Cases
- After completing a task, a user runs /sync to capture progress and decisions.
- Ending a session with /sync saves all context, decisions, patterns, and troubleshooting notes.
- Discovering new patterns triggers a sync entry under CLAUDE-patterns.md.
- Encountering an issue and solving it is logged in CLAUDE-troubleshooting.md during the sync.
- Project audits require running /sync --full to refresh the capability inventory.