Agent Skills: Creating Mods

Creating Mods

Use this skill to create or update trusted Letta Code mod files. Mods are trusted local code that add small composable capabilities through mod APIs, not by importing app internals. Dynamic agent/conversation/workspace/model state is passed as ctx to tool, command, event, and permission callbacks (panels receive live agent/model in their render context); do not read mutable global context for model-callable behavior. Prefer scoped handles (ctx.conversation, ctx.cwd, ctx.agent) and guard optional UI with letta.capabilities.

Capabilities vary by surface — not every surface loads every capability. The TUI/headless host can load tools, commands, events, UI, and providers; the desktop listener loads tools, commands, providers, and tool/turn events, but not panel UI. Always guard each registration on the capabilities its behavior needs.

Choose where the mod file lives

Default to a single mod file unless the user asks for something larger.

| Location | Use when | | --- | --- | | ~/.letta/mods/foo.ts | The behavior should apply to local sessions on this machine. Use this by default. | | $MEMORY_DIR/mods/foo.ts | The behavior should travel with one agent's MemFS/memory. |

Do not create project mods.

Packaging is an upgrade path, not the default authoring path. If the user asks to share, publish, distribute, or use third-party package dependencies, first build a working mod file, then use letta mods package <mod-file> --name <package-name>. Package install/update/download/publish details belong outside this skill.

Choose the right capability

| User wants | Build | | --- | --- | | Agent/model should autonomously call a local capability | Mod tool | | User wants /foo to send a prompt or run local UI logic | Mod command | | Slash command represents a reusable agent workflow | Skill + thin mod command | | Command should work while the main agent is busy | Command with runWhenBusy: true, handled, panel/status, and usually ctx.conversation.fork() | | Show transient output above input | Panel, usually from a command | | Show small persistent state | Status value | | React to app/session lifecycle or transform outbound turns | Event | | Enforce dynamic allow/ask/deny policy for tool calls | Permission overlay | | Add a custom model/API provider for local agents | Provider mod (local agents only) | | Change the bottom statusline appearance | Use customizing-statusline, not this skill |

Default to a tool when the model should decide when to use the capability. Default to a command when the human explicitly invokes it. Compose capabilities when the UX needs it, e.g. command + panel + scoped conversation fork.

Workflow

1. Pick the target scope: harness mod file (~/.letta/mods/) by default, or agent mod file ($MEMORY_DIR/mods/) only when the behavior should travel with this agent.
2. Inspect the relevant mods directory for related files.
3. Preserve unrelated mod code. Prefer a focused new file if merging would be messy.
4. Choose the mod shape and load only the needed recipe:
   • tools: references/tools.md
   • commands: references/commands.md
   • local custom providers: references/providers.md
   • events: references/events.md
   • permissions: references/permissions.md
   • panels/status/capabilities: references/ui.md
   • complex plan-mode composition: references/plan-mode.md
5. For multi-capability or stateful mods, also read references/architecture.md.
6. Write a single-file mod unless the user asks for something larger.
7. Return disposers for registered providers/commands/tools/events, timers, subscriptions, and panels that should close on reload.
8. Do a basic review: valid names, descriptions present, schemas are object schemas, optional capabilities guarded, scoped APIs used, cleanup returned.
9. Tell the user the absolute file path changed and to run /reload. If a mod breaks startup or command handling, recover with letta --no-mods or LETTA_DISABLE_MODS=1 letta.

Core mod shape

export default function activate(letta) {
  const disposers = [];

  if (letta.capabilities.tools) {
    disposers.push(letta.tools.register(/* ... */));
  }

  if (letta.capabilities.commands) {
    disposers.push(letta.commands.register(/* ... */));
  }

  return () => {
    for (const dispose of disposers.reverse()) dispose();
  };
}

Use letta.capabilities for optional behavior:

letta.capabilities.tools
letta.capabilities.commands
letta.capabilities.events.lifecycle
letta.capabilities.events.tools
letta.capabilities.events.turns
letta.capabilities.events.compact
letta.capabilities.events.llm
letta.capabilities.permissions
letta.capabilities.providers
letta.capabilities.ui.panels

Guard each registration on every capability its behavior depends on — not just the one that registers it. Surfaces load different capability subsets, so a registration that relies on another capability (a command that opens UI, emits an event, or calls a provider) must guard on that capability too. Otherwise it is advertised or activated on a host that cannot fulfill it and silently does nothing. Register where the host can actually do the work.

Scoped API model

• In commands and events, use ctx.conversation for conversation operations:
  • ctx.conversation.getHistory() for recent messages
  • ctx.conversation.fork() for independent/background model work
  • forked.sendMessageStream([...]) to stream from a fork
• In tools, use ctx.conversation.getHistory() when the tool needs recent context.
• Use letta.client only for server-specific Letta API calls; do not use it as a substitute for scoped conversation helpers.
• Do not import @/backend, @/cli, or other Letta Code internals from mod files.

Diagnostics

Use letta.diagnostics.report({ message, severity }) sparingly as a debug utility for mod setup/runtime problems an agent should inspect, such as missing required environment variables or failed local configuration. Default severity is "error"; use severity: "warning" only for optional/degraded behavior. Keep messages short and actionable, and do not dump routine logs or large state.

Agents can inspect local mod diagnostics at:

~/.letta/mods/diagnostics/latest.json

Rules

• Do not create project mods.
• Custom provider mods are local-backend/local-agent only. They do not add providers for Constellation/cloud agents.
• Provider mods may run in a provider-only listener context; keep provider registration independent from commands/tools/UI and guard everything else.
• Direct mod files should not assume third-party npm packages are available. Use Node/Bun built-ins unless packaging is explicitly requested.
• Do not do surprising side effects on startup; mods activate on app start and /reload.
• Keep user-facing output short and intentional.
• Prefer Node/Bun standard APIs (node:child_process, node:fs, etc.) for local work.
• For shell execution, prefer execFile/spawn over shell strings.
• Do not use emojis for loading states; use text or spinner-like characters if the user asks for loading UI.
• For runWhenBusy: true, do not return prompt; return handled and own the UI/background work.
• Treat turn_start as powerful trusted code: keep transforms narrow and unsurprising.

Pre-flight checklist for complex mods

Before finishing, verify:

• The mod has one clear owner/file and does not mix unrelated features.
• Command/tool IDs are valid; command overrides of built-ins are intentional, and tool IDs do not collide with built-ins.
• Tool descriptions explain when the model should call them.
• JSON schemas are object schemas with useful descriptions.
• Optional UI/event APIs are capability-guarded.
• Each registration is guarded by every capability its behavior depends on, not just the one that registers it, so it isn't advertised or activated on a surface that can't fulfill it.
• Provider mods are capability-guarded and clearly documented as local-agent only.
• Timers, intervals, event registrations, and panels are cleaned up in a disposer.
• Busy commands return { type: "handled" } quickly and avoid main-conversation sends.
• Conversation work uses ctx.conversation or forked handles, not app internals.
• Local shell/file work is scoped to ctx.cwd unless intentionally global.
• Errors shown to the user are short and actionable.

References

| Reference | Load when | | --- | --- | | references/tools.md | The model should autonomously call a local capability | | references/commands.md | The human should invoke /foo | | references/providers.md | Adding a custom model/API provider for local agents | | references/events.md | Reacting to lifecycle/tool/turn events or transforming turns/tools | | references/permissions.md | Enforcing dynamic tool allow/ask/deny policy before approval/execution | | references/ui.md | Panels (including order-0 statusline and order-1 dreaming indicator) or ui.panels capability guards are involved | | references/plan-mode.md | Recreating plan mode with commands, tools, events, permissions, and local state | | references/analysis-mode.md | Phrase-triggered diagnostic mode with turn reminders (simpler than plan-mode) | | references/architecture.md | Multiple capabilities, local state, cleanup, background model work, or non-trivial composition |