Aide

aide — Claude Code Plugin

A diagnostic assistant for the aide CLI. It diagnoses sandbox issues, context mismatches, and missing credentials by inspecting your local configuration and runtime state. When problems are found, it suggests the safest possible fix first and clearly explains security trade-offs for any broadening change, so you always know what you are allowing before you allow it.

Installation

# Add the aide repo as a marketplace
/plugin marketplace add jskswamy/aide

# Install the plugin
/plugin install aide@jskswamy-aide

# Reload in active session
/reload-plugins

Team Setup

For projects using aide, add to .claude/settings.json:

{
  "extraKnownMarketplaces": {
    "aide-plugins": {
      "source": {
        "source": "github",
        "repo": "jskswamy/aide"
      }
    }
  }
}

Commands

Command	Purpose
/aide	Quick status — shows current context, warnings, routes to other commands
/aide setup	Guided setup — wraps aide init or aide setup
/aide doctor	Full diagnostic — validates config, sandbox, context, secrets
/aide sandbox	Sandbox diagnostics — policy review, guard management, path tuning
/aide context	Context management — match resolution, add/modify/rename contexts
/aide secrets	Secret management — list, create, edit, rotate encrypted secrets
/aide env	Environment variables — list, set, wire from secrets
/aide config	Config review — validation, hardening suggestions, optimization tips
/aide agents	Agent management — list, add, remove, check binary availability
/aide use	Quick bind — bind directory to agent/context with guided options

Auto-Triggering Skills

Skill	Activates When
sandbox-doctor	"permission denied", "agent hanging", "can't write to"
context-doctor	"wrong agent", "wrong context", "why is it using"
secrets-doctor	"can't access API", "missing key", "authentication failed"
setup-guide	"set up aide", "configure aide", "initialize aide"
config-review	"review config", "is my config correct", "optimize aide"

Configuration

User preferences in .claude/aide-plugin.local.md:

---
session_start:
  show_warnings: true
  show_tips: true
---

How It Works

• Every command follows: Observe, Diagnose, Recommend, Apply
• Flags and subcommands are discovered at runtime via aide --help (never hardcoded)
• Fixes are classified as Safe or Broadening (security trade-offs explained for broadening changes)
• Proactive tips offered after each interaction