Skill sync

Skill light-merlin-dark/skill-sync/skills/skill-sync

Sync local repo-backed agent skills across Codex, Claude Code, Cursor, Gemini, Hermes, and more. Safe symlink sync, drift checks, backups, and restore.

Install

npx -y skills add light-merlin-dark/skill-sync --skill skill-sync

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Sync local repo-backed agent skills across installed harnesses such as Codex, Claude Code, Cursor, Gemini, Hermes, Grok, and related tools. Use when a user wants one source of truth for local SKILL.md files, needs drift checks or backup/restore for harness skill roots, or wants to inspect which harnesses and skills are currently detected.

SKILL.md

3.0 KB, 615 tokens by cl100k_base, as published. Nobody here has run it

Skill Sync

Use skill-sync as the default interface for local skill-harness maintenance.

Core Workflow

Inspect harness detection and discovered skill sources.
Run a doctor pass before making changes.
Create a backup before risky cleanup or restore work.
Execute or restore.
Verify the resulting symlinks or restored content.

Start with:

skill-sync harnesses
skill-sync sources
skill-sync doctor
skill-sync doctor --verbose

Apply changes:

skill-sync execute
skill-sync sync
skill-sync execute --skill <slug>   # preferred for one project/skill

Or explicitly:

skill-sync execute

Backup Workflow

Create a backup:

skill-sync backup create

List backups:

skill-sync backup list

Dry-run a restore:

skill-sync backup restore <backup-id> --dry-run

Restore:

skill-sync backup restore <backup-id>

Agent-Friendly Usage

Use JSON when the output will be consumed by another tool or agent:

skill-sync doctor --json
skill-sync sources --json
skill-sync harnesses --json
skill-sync execute --json

Bare skill-sync prints a high-signal landing/help view. Default human output is concise. Add --verbose when you need the full per-entry plan and orphan listing.

Safety Rules

Prefer doctor before execute.
For a project-local release, use doctor --skill <slug> then execute --skill <slug>. Targeted mode never prunes unrelated managed entries and does not inherit unrelated source conflicts.
execute applies all non-conflicting changes by default. Conflicting entries are skipped but still reported (exit code 3 signals remaining issues).
Divergent harness-native skills with disjoint local-only destinations are valid and do not block global execution.
If doctor reports a conflict due to an existing unmanaged install (common case: a skill folder already exists in a harness root like ~/.hermes/skills/<skill>), resolve by either:
- removing the unmanaged directory/file and re-running execute, or
- restoring via skill-sync backup restore <backup-id>. Do not leave mixed symlink + real directories behind.
Use --home for isolated testing against a fake home directory.
Use --projects-root when you need to constrain discovery to a specific source tree.
Project-root sources are authoritative. Harness-installed skills act as fallback sources when no project-root source exists for the same slug.
If a harness-native skill should not fan out globally, add skill-sync-scope: local-only to its frontmatter. Use skill-sync-install-on: [harness-a, harness-b] for explicit multi-harness targeting.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.