Agent Skills: AIDA Dispatch

AIDA Dispatch

Routes /aida commands to appropriate action handlers, managing AIDA's configuration, diagnostics, feedback systems, and extension management (agents, skills, plugins).

Activation

This skill activates when:

• User invokes /aida command with any action
• AIDA functionality is needed (configuration, status checks, diagnostics)
• Command routing and execution orchestration is required

Command Routing

When this skill activates, check the <command-args> tag to determine which action to route:

Diagnostic Commands

For status, doctor, or upgrade commands:

• Read references/diagnostics.md for execution workflow
• These are non-interactive commands that execute Python scripts directly

Configuration Commands

For config command (alone):

• Read references/config.md for YAML-based configuration workflow
• This is an interactive command with:
  • Dynamic menu generation based on installation state
  • Auto-detection of project facts (saved to YAML)
  • Minimal questions (0-3) for unknown preferences only
  • Automatic skill generation from YAML config

For config validate command:

• Read references/validate.md for the workflow
• This is a non-interactive CI-friendly health check
• Run ~/.aida/venv/bin/python3 {base_directory}/scripts/validate.py
• Exits 0 on success, 1 on failure — suitable for make / CI gates
• Pass --json for machine-consumable output

Permissions Commands

For config permissions or permissions commands:

• Invoke the permissions skill to handle these operations
• The skill handles interactive permission setup and audit mode

Process:

1. Parse the command to detect:

   • --audit flag for audit mode
   • No flags for interactive setup

2. Invoke permissions skill with the parsed context

Examples:

/aida config permissions           → permissions skill
/aida config permissions --audit   → permissions skill (audit)

Feedback Commands

For feedback, bug, or feature-request commands:

• Read references/feedback.md for feedback collection workflow
• These are interactive commands that collect and submit user input

Help Command

For help or no arguments:

• Display the help text inline (see Help Text section below)
• Read the plugin version from {base_directory}/../../.claude-plugin/plugin.json (the "version" field) and include it in the footer
• No additional reference files need to be loaded

About Command

For about:

• Read {base_directory}/../../.claude-plugin/plugin.json
• Resolve the installed path from {base_directory}/../..
• Display plugin metadata:

**AIDA Core Plugin**

- **Version:** {version from plugin.json}
- **Author:** {author.name from plugin.json}
- **Repository:** {repository from plugin.json}
- **Installed at:** {resolved path to plugin root}

Agent Management Commands

For agent commands:

• Invoke the agent-manager skill to handle these operations
• Pass the full command arguments to the skill
• The skill handles create, validate, version, and list operations

Process:

1. Parse the command to extract:

   • Operation: create, validate, version, list
   • Arguments: name, description, options

2. Invoke agent-manager skill with the parsed context

Examples:

/aida agent create "description"     → agent-manager skill
/aida agent validate --all           → agent-manager skill
/aida agent version my-agent patch   → agent-manager skill
/aida agent list                     → agent-manager skill

Skill Management Commands

For skill commands:

• Invoke the skill-manager skill to handle these operations
• Pass the full command arguments to the skill
• The skill handles create, validate, version, and list operations

Process:

1. Parse the command to extract:

   • Operation: create, validate, version, list
   • Arguments: name, description, options

2. Invoke skill-manager skill with the parsed context

Examples:

/aida skill create "description"     → skill-manager skill
/aida skill validate --all           → skill-manager skill
/aida skill version my-skill patch   → skill-manager skill
/aida skill list                     → skill-manager skill

Plugin Management Commands

For plugin commands (including plugin scaffold):

• Invoke the plugin-manager skill to handle these operations
• Pass the full command arguments to the skill
• The skill handles create, validate, version, list, scaffold, update, deps, and agents operations
• Scaffold creates a NEW plugin project (not an extension inside an existing project)
• Update scans an existing plugin and patches it to current standards
• Deps reports declared plugin dependencies + their satisfied / missing / wrong-version status (#20)

Process:

1. Parse the command to extract:

   • Operation: create, validate, version, list, scaffold, update, deps
   • Arguments: name, description, options

2. Invoke plugin-manager skill with the parsed context

Examples:

/aida plugin create "description"        → plugin-manager skill
/aida plugin validate --all              → plugin-manager skill
/aida plugin list                        → plugin-manager skill
/aida plugin scaffold "my-new-plugin"    → plugin-manager skill
/aida plugin scaffold                    → plugin-manager skill (will ask)
/aida plugin update "/path/to/plugin"   → plugin-manager skill
/aida plugin deps                        → plugin-manager skill (cwd)
/aida plugin deps "/path/to/plugin"     → plugin-manager skill
/aida plugin agents                      → plugin-manager skill (cwd)
/aida plugin agents "/path/to/project"  → plugin-manager skill

Hook Management Commands

For hook commands:

• Invoke the hook-manager skill to handle these operations
• Pass the full command arguments to the skill
• Hooks are settings.json config, not files
• The skill handles list, add, remove, and validate operations

Process:

1. Parse the command to extract:

   • Operation: list, add, remove, validate
   • Arguments: event, matcher, command, scope

2. Invoke hook-manager skill with the parsed context

Examples:

/aida hook list                      → hook-manager skill
/aida hook add "auto-format"         → hook-manager skill
/aida hook remove my-hook            → hook-manager skill
/aida hook validate                  → hook-manager skill

Expert Registry Commands

For expert commands:

• Invoke the expert-registry skill to handle operations
• Pass the full command arguments to the skill
• The skill handles list, configure, panels, and panel operations

Process:

1. Parse the command to extract:

   • Operation: list, list configure, panel list, panel create, panel remove
   • Arguments: panel name (for create/remove)

2. Invoke expert-registry skill with the parsed context

Examples:

/aida expert list                   → expert-registry skill
/aida expert list configure        → expert-registry skill
/aida expert panel list            → expert-registry skill
/aida expert panel create review   → expert-registry skill
/aida expert panel remove review   → expert-registry skill

Knowledge Commands

/aida knowledge commands split across two skills:

• knowledge-sync skill owns the mechanic — sync, status, discover. Deterministic: reads source declarations, fetches content, walks spider roots.
• knowledge-curator skill owns the policy workflow — curate, review. LLM-orchestrated: reasons about which discovered URLs are worth using and persists decisions.

Process:

1. Parse the command to extract:

   • Operation: sync, status, discover, curate, review
   • Argument: agent name (and any flags)

2. Route to the appropriate skill:

| Command | Skill | Script | | ------- | ----- | ------ | | sync <agent> | knowledge-sync | sync.py --agent <name> | | status <agent> | knowledge-sync | sync.py --agent <name> --dry-run | | discover <agent> | knowledge-sync | discover.py --agent <name> | | audit <agent> | knowledge-sync | audit.py --agent <name> | | promote <agent> --url X --file F --section S | knowledge-sync | promote.py --agent <name> --url X --file F --section S | | regenerate-md <agent> | knowledge-sync | regenerate_md.py --agent <name> | | curate <agent> | knowledge-curator | follow SKILL.md workflow | | review <agent> | knowledge-curator | follow SKILL.md workflow |

Script paths are resolved as ~/.aida/venv/bin/python3 {base_directory}/../<skill>/scripts/<file>.py.

Examples:

/aida knowledge sync <agent>           → knowledge-sync (sync.py)
/aida knowledge status <agent>         → knowledge-sync (--dry-run)
/aida knowledge discover <agent>       → knowledge-sync (discover.py)
/aida knowledge audit <agent>          → knowledge-sync (audit.py)
/aida knowledge promote <agent> ...    → knowledge-sync (promote.py)
/aida knowledge regenerate-md <agent>  → knowledge-sync (regenerate_md.py)
/aida knowledge curate <agent>         → knowledge-curator (workflow)
/aida knowledge review <agent>         → knowledge-curator (workflow)

Memento Commands

For memento commands:

• Invoke the memento skill to handle these operations
• Pass the full command arguments to the skill
• The skill handles create, read, list, update, complete, and remove operations

Process:

1. Parse the command to extract:

   • Operation: create, read, list, update, complete, remove
   • Arguments: description, slug, source, filter options

2. Invoke memento skill with the parsed context

Examples:

/aida memento create "description"   → memento skill
/aida memento create from-pr         → memento skill (source=from-pr)
/aida memento create from-changes    → memento skill (source=from-changes)
/aida memento read my-memento        → memento skill
/aida memento list                   → memento skill
/aida memento list --filter active   → memento skill
/aida memento list --all             → memento skill (all_projects=true)
/aida memento list --project foo     → memento skill (project_filter=foo)
/aida memento update my-memento      → memento skill
/aida memento complete my-memento    → memento skill
/aida memento remove my-memento      → memento skill

CLAUDE.md Management Commands

For claude commands:

• Invoke the claude-md-manager skill to handle these operations
• Pass the full command arguments to the skill
• The skill handles create, optimize, validate, and list operations

Process:

1. Parse the command to extract:

   • Operation: create, optimize, validate, list
   • Arguments: scope (project/user/plugin), path, options

2. Invoke claude-md-manager skill with the parsed context

Examples:

/aida claude create                  → claude-md-manager skill (auto-detect scope)
/aida claude create --scope project  → claude-md-manager skill (scope=project)
/aida claude create --scope user     → claude-md-manager skill (scope=user)
/aida claude optimize                → claude-md-manager skill (audit current)
/aida claude optimize ./CLAUDE.md    → claude-md-manager skill (audit specific)
/aida claude validate                → claude-md-manager skill
/aida claude list                    → claude-md-manager skill

Path Resolution

Base Directory: Provided when skill loads via <command-message> tags containing the skill base directory.

Script Execution: Construct full paths from base directory:

{base_directory}/scripts/status.py
{base_directory}/scripts/doctor.py
{base_directory}/scripts/upgrade.py
{base_directory}/scripts/detect.py
{base_directory}/scripts/configure.py
{base_directory}/scripts/install.py
{base_directory}/scripts/feedback.py

Reference Loading: Reference files are located in references/ subdirectory:

{base_directory}/references/diagnostics.md
{base_directory}/references/config.md
{base_directory}/references/feedback.md

Help Text

When displaying help (for help command or no arguments), show:

## Available AIDA Commands

### Configuration & Setup
- `/aida config` - Configure AIDA settings (global or project-level)
- `/aida config validate` - Non-interactive CI-friendly health check
- `/aida config permissions` - Configure Claude Code permissions from plugin recommendations
- `/aida status` - Check AIDA installation and configuration status
- `/aida doctor` - Run diagnostics to troubleshoot AIDA issues

### Maintenance
- `/aida upgrade` - Check for and install AIDA updates

### Feedback & Support
- `/aida feedback` - Submit feedback about AIDA
- `/aida bug` - Report a bug in AIDA
- `/aida feature-request` - Request a new AIDA feature

### Extension Management
- `/aida agent [create|validate|version|list]` - Manage agents
- `/aida skill [create|validate|version|list]` - Manage skills
- `/aida plugin [scaffold|create|validate|version|list|update]` - Manage plugins
- `/aida hook [list|add|remove|validate]` - Manage hooks (settings.json)

### Expert Registry
- `/aida expert list` - List available experts and activation status
- `/aida expert list configure` - Select active experts (project or global)
- `/aida expert panel list` - Show named panel compositions
- `/aida expert panel create <name>` - Create a named expert panel
- `/aida expert panel remove <name>` - Remove a named panel

### Knowledge
- `/aida knowledge sync <agent>` - Sync an agent's knowledge from declared sources + in-use decisions
- `/aida knowledge status <agent>` - Dry-run; report what would change
- `/aida knowledge discover <agent>` - Spider configured roots; record new URLs as pending decisions
- `/aida knowledge audit <agent>` - Drift report (pending count, stale verdicts, never-synced URLs)
- `/aida knowledge promote <agent> --url X --file F --section S` - Manually mark a URL in-use (skip the LLM curator)
- `/aida knowledge regenerate-md <agent>` - Repair: rebuild decisions.md from decisions.json
- `/aida knowledge curate <agent>` - LLM workflow: decide pending URLs into in-use or rejected
- `/aida knowledge review <agent>` - Interactive: human confirms or overrides curator decisions

### Session Persistence
- `/aida memento create "description"` - Save current work context
- `/aida memento create from-pr` - Create from current PR
- `/aida memento create from-changes` - Create from file changes
- `/aida memento read <slug>` - Load memento into context
- `/aida memento list` - List active mementos
- `/aida memento list --all` - List mementos from all projects
- `/aida memento list --project <name>` - List mementos for specific project
- `/aida memento update <slug>` - Update memento sections
- `/aida memento complete <slug>` - Archive completed memento

### CLAUDE.md Management
- `/aida claude create` - Create CLAUDE.md with auto-detection
- `/aida claude create --scope user` - Create user-level CLAUDE.md
- `/aida claude optimize` - Full audit with scoring and findings
- `/aida claude validate` - Validate CLAUDE.md structure
- `/aida claude list` - List all CLAUDE.md files in hierarchy

### Info
- `/aida help` or `/aida` - Show this help message
- `/aida about` - Show plugin version and metadata
- `/aida status` - Show installation status and version

## Getting Started

If you haven't configured AIDA yet: `/aida config`
To check if AIDA is working: `/aida status`
If you encounter issues: `/aida doctor`
To create an agent: `/aida agent create "description"`
To save work context: `/aida memento create "description"`
To optimize your CLAUDE.md: `/aida claude optimize`
To list hooks: `/aida hook list`
To add a hook: `/aida hook add "auto-format on write"`

---
aida-core v{version from plugin.json}

Resources

scripts/

Executable Python scripts for AIDA operations:

• status.py - Display current installation and configuration state
• doctor.py - Run health checks and diagnostics
• upgrade.py - Check for and install updates
• detect.py - Detect current installation state (used by config)
• configure.py - Interactive project configuration
• install.py - Global installation setup
• feedback.py - Submit feedback, bugs, and feature requests
• utils/ - Shared utilities (paths, files, version, questionnaire, etc.)

references/

Detailed workflow guides loaded as needed:

• diagnostics.md - Workflow for status/doctor/upgrade commands
• config.md - YAML-based configuration flow with auto-detection
• feedback.md - Feedback collection and submission workflow
• config-driven-approach.md - Architecture documentation for config system
• project-facts.md - Comprehensive taxonomy of detectable project facts