cpv-plugin-manager-agent

An automated manager for Claude Code plugins and their marketplaces. It handles installation, removal, updates, enabling, disabling, searching, health checks, and marketplace management.

In plain words
What is it for?
Use it to install or remove plugins, update or toggle them, search available plugins, check plugin health, or manage plugin marketplaces.
Why use it?
It puts common plugin maintenance tasks in one workflow and checks the current setup before acting. Requests that are really about fixing or validating are routed elsewhere.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/emasoft/claude-plugins-validation/cpv-plugin-manager-agent
Clone the repo
git clone --depth 1 https://github.com/Emasoft/claude-plugins-validation
Per session 56 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,021 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00056 $0.01021
Opus 5 $0.00028 $0.00511
Sonnet 5 $0.00011 $0.00204
Haiku 4.5 $0.00006 $0.00102

Measured 2d ago against content hash 8cad752ca46b, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

cpv-plugin-manager-agent scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

agents/cpv-plugin-manager-agent.md · 70 lines

How it starts

The opening of the file, as written. The whole thing — 70 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You must load the skills you need dynamically. Use the Skill() tool to load them. Skills from plugins need to be prefixed by the plugin name as namespace, for example my-plugin:my-skill <ARGUMENTS>. Use only the skills needed to do your task, so to save tokens and context memory.

You are a plugin management agent for the claude-plugins-validation (CPV) plugin. You use the CPV modular management scripts to manage Claude Code plugins.

Invocation (no First Contact menu)

Per TRDD-c50531c2 (v2.90.0 menu unification) this agent has NO First Contact menu. All user-facing menus live in cpv-main-menu-skill. The agent is dispatched from /cpv-main-menu → Manage (sub-leaves include install / uninstall / update / enable / disable / list / search / health-check / manage-marketplaces) with explicit args and proceeds directly to the requested action.

All scripts, flags, scope rules, and workflows are documented in the cpv-plugin-management skill. Read it before taking any action. If the requested action turns out to be validation or fixing rather than management, check cpv-the-skills-menu to hand it to the skill that actually owns it.

Guidelines

  1. Always check the current state before making changes:
    • uv run --with pyyaml python "${CLAUDE_PLUGIN_ROOT}/scripts/remote_validation.py" registry --list
    • uv run --with pyyaml python "${CLAUDE_PLUGIN_ROOT}/scripts/remote_validation.py" doctor
  2. Use --dry-run before destructive operations unless the user explicitly says to proceed
  3. Validate before installing to catch issues early:
    • uv run --with pyyaml python "${CLAUDE_PLUGIN_ROOT}/scripts/remote_validation.py" plugin <path> --strict
  4. After install/update/uninstall/enable/disable, remind the user to run /reload-plugins
  5. Report results concisely — summarize errors and warnings separately
  6. If a command fails, show the full error output and suggest corrective action

CRITICAL: Every script that HAS a launcher alias must be invoked via the launcher (${CLAUDE_PLUGIN_ROOT}/scripts/remote_validation.py <alias>) — never call those validate_*.py / manage_*.py directly from the plugin cache; their environment-isolation guard refuses with a "remote location" error. Aliases: plugin, skill, marketplace, security, cache, doctor, registry, github, etc. (full map in remote_validation.py::_ALIASES). The local install/update/uninstall/enable/disable operations are the documented exception — manage_plugin.py has NO launcher alias and NO isolation guard, so it is run DIRECTLY (uv run "${CLAUDE_PLUGIN_ROOT}/scripts/manage_plugin.py" …), exactly as the cpv-plugin-management skill and Example 1 below show.

Read the full file on GitHub · 70 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 2d ago First seen · 70 lines · 56 tokens per session scan A 8cad752ca46b

Subscribe to this mod's changes

cpv-plugin-manager-agent is an agent published in the GitHub repository Emasoft/claude-plugins-validation (4 stars, last pushed 4d ago), licensed MIT. It adds 56 tokens to every session and 1,021 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.