skills.rest
letta-ai
letta-aiOFFICIAL
Explore all skills by this author →
Website𝕏@letta_aiGitHub

self-configuration

Inspect and modify Letta Code agent memory, model, permissions, and runtime settings.

3.2k|385|Updated Oct 25, 2025
Search Tags:#system prompt#CLI#compaction#Letta#Permissions#model settings#Agent configuration
One-click install
npx skills add https://github.com/letta-ai/letta-code --skill self-configuration
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: self-configuration
Source: https://github.com/letta-ai/letta-code/tree/main/src/skills/builtin/self-configuration
Command: npx skills add https://github.com/letta-ai/letta-code --skill self-configuration

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Changing how a Letta Code agent behaves requires choosing the right configuration layer—memory files, server agent fields, conversation overrides, local settings, mods, channels, or schedules—and a wrong-layer change can clobber system prompts, brick the agent, or leak secrets.

Core Features & Use Cases

  • Layered configuration routing: Decision tables map each request (rename, model change, permission rule, schedule, channel setup) to the correct layer, with a safe inspect-then-patch workflow and rollback guidance.
  • Guarded server updates: Helper scripts patch agent/conversation model, context window, model settings, compaction prompts, and system prompts with dry-run previews, cross-agent mismatch rejection, and confirmation gates for dangerous writes.
  • Local settings and permissions management: Scripts report merged user/project/local settings without printing secrets and add allow/deny/ask/alwaysAsk permission rules atomically.
  • Use Case: A user asks to switch their agent to a new model with a larger context window; the skill dry-runs a conversation-scoped patch first, verifies the effective state, then persists the agent-level default.

Quick Start

Ask the agent to show its current configuration and change its default model to a specific provider handle with a dry run first.

Frequently Asked Questions about self-configuration

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I change the model or context window of a Letta Code agent?▼

Use the update-agent-settings.ts script with --target agent, --model, and --context-window-limit, starting with --dry-run to preview the patch. Test risky changes on the conversation scope first, then persist them as agent defaults.

How do I rename a Letta Code agent?▼

Patch the agent name through the server API using update-agent-settings.ts with --target agent and --name, since editing persona memory alone does not change the actual agent name. Then update any memory file that states the old name so both agree.

What is the difference between agent-level and conversation-level settings?▼

Agent-level fields set persistent defaults like model, system prompt, and compaction, while conversation-level fields hold temporary overrides for one conversation. Do not infer one from the other; inspect both scopes when diagnosing model or context differences.

How do I add permission rules to Letta Code settings?▼

Run add_permission.py with --rule, --type (allow, deny, ask, or alwaysAsk), and --scope (user, project, or local). User-scope writes require --confirm-user-scope, and --dry-run previews the change without writing.

Why is changing the system prompt risky for a Letta agent?▼

A bad replacement system prompt can self-brick the agent by preventing it from completing turns. Live writes require --confirm-system-replacement, and ordinary behavior changes should edit memory files instead of replacing the compiled prompt.

Can I recover an agent broken by a bad model or prompt change?▼

Yes, recover out of band from another shell or client using the CLI or API with explicit agent IDs, since the broken model cannot repair itself. The helper scripts permit explicit IDs when the environment ID is absent for this recovery path.

Related Skills

Explore More Software Engineering Skills→
cpfiffercpfiffer
23

using-letta-api

Manage agents and subagents via the Letta Python client.

Community
Intermediate
stevengonsalvezstevengonsalvez

add-compact

Add a /compact command to manually compact agent conversation context.

Community
Advanced
t0labt0lab
2

find-skills

Discover and install agent skills via the Skills CLI.

Official
Intermediate
Z43LZ43L

1password

Read and inject credentials from the 1Password CLI into commands and templates.

Community
Intermediate