Install and Manage AI Skills

This skill installs, validates, registers, and verifies new Claude Code skills through a 6-layer redundancy validation system: detect_skills.py confirms SKILL.md exists with valid frontmatter, validate_skill.py runs 10 deep checks, install_skill.py validates pre-install (conflicts, permissions, disk space, version) and post-install (files copied correctly), scan_registry.py confirms the skill appears in the registry with deduplication, and package_skill.py validates the resulting ZIP (no backslashes, non-empty, integrity-checked). The main installer runs 11 steps: resolve the source path, validate the skill, determine its name (from frontmatter or an override, comparing versions), check for destination conflicts, take a timestamped backup before any overwrite, copy via a staging area with hash validation before the final move, register the SKILL.md into .claude/skills/<name>/ for Claude Code CLI, update the registry with name-based deduplication, verify the installation with 5 checks, package a validated ZIP for Claude.ai web/desktop upload, and log the operation with automatic rotation - covering both the CLI and web/desktop skill surfaces, which are separate systems. It supports ten scenarios via CLI flags: installing after skill-creator finishes, installing a specific skill (with optional name override or force overwrite), dry-run simulation, detecting and optionally auto-installing pending skills from known locations (Desktop, Downloads, Temp, workspaces), uninstalling with automatic backup, a health check across all skills (frontmatter, registration, registry, duplicates) with an auto-repair mode, rollback to the most recent backup with automatic re-registration, reinstall-all for mass re-registration after bulk changes, a rich status dashboard, and an operation history log. Security measures include automatic timestamped backups before overwrites, a staging area with hash validation to minimize corruption, idempotency detection (running twice with identical source doesn't duplicate), blocking installation of forbidden files (.env, *.key, .pem, credentials.), rotating logs (last 500 entries kept), a 5-backup-per-skill retention limit with automatic cleanup, anti-recursion exclusion of the tool's own backup/staging subdirectories, case-insensitive registry deduplication, ZIP integrity validation, and version comparison (detecting upgrade/downgrade/same) before overwriting. It integrates with an orchestrator: auto-detected by scan_registry.py and matched by match_skills.py when a user mentions installation keywords, with no manual configuration needed, and is triggered automatically after skill-creator finishes producing a new skill per global CLAUDE.md instructions.

Use it when a user mentions installing, registering, or adding a new skill to the ecosystem, or right after a skill-creator tool finishes producing a skill. Do not use it for tasks unrelated to skill installation, when a simpler specific tool would handle the request, or for general-purpose assistance outside its domain.

Inputs: the source path of a skill to install (or --detect to scan known locations), plus optional flags for naming, forcing, dry-run, uninstall, rollback, health-check, or status.
Outputs: an installed and registered skill (CLI + optionally packaged ZIP for Claude.ai), a JSON validation report (valid/checks/warnings/errors), a rich status dashboard, or an operation history log, depending on the invoked scenario.

python install_skill.py --source "<skill-path>" --force
python install_skill.py --health --repair
python install_skill.py --rollback "skill-name"

Teams managing a growing library of Claude Code skills who need reliable, auditable installation, health-checking, backup/rollback, and dual-surface (CLI + Claude.ai) packaging rather than manual file copying.