Agent Call
Plugincall another agent such as cursor-agent, codex, claude to perform a task.
This plugin provides three skills: `call-claude`, `call-codex`, and `call-cursor-agent`, each designed to invoke a specific AI agent's CLI with a user-provided task description. It assumes the target CLIs are installed and configured on the user's system.
Practical Utility
- info:Unique selling propositionThe extension appears to be a thin wrapper around CLI commands for different agents, with limited added value beyond direct invocation.
- warning:Production readinessThe skills are basic wrappers and lack comprehensive error handling or configuration options, which limits their production readiness.
Documentation
- warning:Configuration & parameter referenceThe SKILL.md files mention parameters like '{task_description}' and mention timeout settings but do not document default values or explicit precedence orders.
Security
- info:Secret ManagementThe SKILL.md files show placeholder commands that use CLI arguments which could potentially expose secrets if not handled carefully by the underlying CLIs.
- warning:InjectionThe `codex exec --sandbox danger-full-access` command and the `claude -p --dangerously-skip-permissions` command appear to bypass security sandboxing without explicit justification or warnings, posing an injection risk.
- warning:Transitive Supply-Chain GrenadesThe use of commands like `which` and direct CLI execution without clear sandboxing or input validation could allow for transitive supply-chain attacks if the underlying tools are compromised.
- warning:Sandbox IsolationThe `codex exec --sandbox danger-full-access` command and the `claude -p --dangerously-skip-permissions` command explicitly bypass sandbox protections, indicating a potential for unintended side effects or security risks.
- warning:Data ExfiltrationThe commands used (e.g., `claude -p`, `codex exec --sandbox danger-full-access`) might allow for unredacted data exfiltration if the underlying tools are not properly configured or if the '{task_description}' includes sensitive information.
- warning:Unguarded Destructive OperationsThe 'codex exec --sandbox danger-full-access' and 'claude -p --dangerously-skip-permissions' commands suggest potentially destructive operations are not adequately guarded or confirmed by default.
Hooks
- warning:Opaque code executionThe skills rely on executing external CLI commands (`claude`, `codex`, `cursor-agent`), which are opaque code execution mechanisms if their internal workings are not fully understood or trusted.
Portability
- warning:Structural AssumptionThe skills assume the presence and correct installation of external CLIs ('claude', 'codex', 'cursor-agent') and that they are in the system's PATH, which might not be true for all user environments.
- warning:Runtime stabilityThe skills assume the availability and specific command-line interfaces of 'claude', 'codex', and 'cursor-agent', which may not be present or configured identically on all systems, potentially leading to silent failures or crashes.
Code Execution
- warning:ValidationThe '{task_description}' parameter is passed directly to CLI commands without explicit validation or sanitization within the skill scripts, potentially allowing for injection if the input is malicious.
- warning:Error HandlingThe skills check for the existence of the CLIs but do not implement robust error handling for cases where the CLI is not found, or if the CLI commands themselves fail, potentially leading to silent failures or opaque errors.
Compliance
- warning:GDPRThe plugin directly invokes external agents with user-provided task descriptions, which could contain personal data. Without explicit sanitization or user consent prompts, this could lead to unintended data submission to third parties.
Invocation
- warning:Name collisionsThe skill names ('call-claude', 'call-codex', 'call-cursor-agent') could potentially collide with Claude Code built-in commands or other installed skills if not properly namespaced by the router.
Pemasangan
Tambah marketplace dahulu
/plugin marketplace add dotneet/claude-code-marketplace/plugin install agent-call@dotneet-marketplaceSambungan serupa
Agent SDK Development Plugin
98Development kit for working with the Claude Agent SDK
Plugin Master
99Complete Claude Code plugin development system. PROACTIVELY activate when users want to: (1) Create/build plugins with 2025 features, (2) Add skills/commands/agents/hooks, (3) Validate plugin structure, (4) Publish to marketplace, (5) Get plugin development guidance. Provides: agent-first design patterns, progressive disclosure skills, hook automation, MCP integration, marketplace publishing. Includes plugin-expert agent and validation utilities.
Plugin Development Toolkit
99Comprehensive toolkit for developing Claude Code plugins. Includes 7 expert skills covering hooks, MCP integration, commands, agents, and best practices. AI-assisted plugin creation and validation.
PR Review Toolkit
98Comprehensive PR review agents specializing in comments, tests, error handling, type design, code quality, and code simplification