Skill

agentic-share

Imports/exports reusable assets (skills, templates, agents) between projects and v0.2 Claude Code plugin repositories, with sanitization for secrets/paths and safety checks.

Markdown

Git

Bash

developer-tools

automation

npx claudepluginhub waterplanai/agentic-config --plugin ac-tools

Tool Access

This skill is limited to using the following tools:

ReadWriteEditBashGlobGrep

Preview

Core logic for moving reusable assets between projects and this repository.

SKILL.md

Similar Skills

agentic-export

Exports skills, templates, or agents from a project to a repository's v0.2 plugin architecture via /agentic-export. Supports dry-run, plugin targeting, and pre-flight checks.

7 tools

ac-tools

cm-skill-share

Safely packages, reviews, exports, and imports CodyMaster skill folders across machines, repos, or teams. Includes checklists for validation, companion files, and discovery updates.

brain-share

Shares local Claude Code skills, agents, or rules to shared brain namespace by copying to git repo, confirming with user, and committing/pushing. Use /brain-share skill my-skill.md for team sync.

claude-brain-sync

Stats

Parent Repo Stars28

Parent Repo Forks6

Last CommitMar 6, 2026

Actions

View Source View Plugin View on GitHub View README

Help us improve

Share bugs, ideas, or general feedback.

Agentic Asset Share

Core logic for moving reusable assets between projects and this repository.

Arguments

mode: import | export
asset_type: skill | template | agent
source: source path or asset name
target: optional destination name/path
options: optional flags (--plugin, --force, --dry-run)

Parsed from: $ARGUMENTS

Canonical Locations (v0.2)

Type	Project Source	Repository Target
skill	`.claude/skills/<name>/` or explicit skill dir	`plugins/<plugin>/skills/<name>/`
template	`templates/<name>/`	`templates/<name>/`
agent	explicit agent markdown path	`plugins/ac-workflow/agents/<name>.md`

Notes:

There is no command asset type in v0.2 (skills-only surface).
For skills, --plugin is required unless destination can be inferred.

Execution Flow

Parse mode, type, source, target, options.
Validate mode (import or export).
Validate asset type (skill, template, agent).
Validate repository root:
- must contain .claude-plugin/marketplace.json
- must contain plugins/
Resolve source and destination paths by type.
Read source content (single file or directory tree).
Apply sanitization:
- scrub absolute home paths (/Users/..., /home/...) -> {USER_HOME}
- scrub secrets/tokens (api_key, token, secret, bearer values) -> {REDACTED}
Validate sanitized output:
- block if unresolved secrets remain
- warn on environment-specific hardcoded values
Write output (or preview with --dry-run).
Report summary and next steps.

Safety Rules

Never overwrite existing targets without --force or explicit confirmation.
Never proceed when unresolved secrets remain after sanitization.
On any failure, abort with a clear error and leave no partial file writes.

Commit Guidance (Optional)

After successful import/export, offer a commit for only touched asset paths.

Conventional message pattern:

feat(assets): import <asset_type> '<asset_name>'

(or export accordingly)

Error Cases

Error	Action
Invalid mode/type	Abort with valid choices
Repository root not detected	Abort and request running from repo root
Source missing	Abort and print source path
Sanitization blockers remain	Abort and print blocker lines
Target exists without force	Ask for overwrite confirmation