migrating-from-codex-and-claude-code_skill

This skill helps you bootstrap agent memory by discovering and searching Claude Code and Codex histories to surface context.
  • TypeScript

1.7k

GitHub Stars

1

Bundled Files

2 months ago

Catalog Refreshed

4 months ago

First Indexed

Readme & install

Copy the install command, review bundled files from the catalogue, and read any extended description pulled from the listing source.

Installation

Preview and clipboard use veilstrat where the catalogue uses aiagentskills.

npx veilstrat add skill letta-ai/letta-code --skill migrating-from-codex-and-claude-code

  • SKILL.md6.0 KB

Overview

This skill helps discover, search, and extract useful information from historical Claude Code and OpenAI Codex CLI conversations stored on the local machine. It gives scripts and commands to detect available history, list sessions, search prompts, and render session content so you can bootstrap agent memory or inspect past coding patterns. The focus is practical: find project context, user preferences, and tool usage from past sessions quickly.

How this skill works

The skill inspects local data directories (~/.claude/ and ~/.codex/) and reads JSONL session and history files. It provides ready-to-run shell scripts that detect available data, list per-project sessions, search across history by keyword, and pretty-print individual session files. Where session files are missing, the tools automatically fall back to global history.jsonl with field-format adjustments for Claude vs Codex.

When to use it

  • During agent initialization to bootstrap memory with project context
  • When a user asks about what they worked on in previous coding sessions
  • To recover project-specific prompts, settings, or session summaries
  • When you need to learn a user’s tooling and command patterns
  • To search for past discussions, decisions, or debugging steps

Best practices

  • Run detect.sh first to understand what history exists before searching
  • Encode project paths for Claude (replace '/' with '-') when listing project sessions
  • Prefer session files for full context; fall back to history.jsonl when sessions are absent
  • Adjust jq queries for Codex (.ts/.text) versus Claude (.timestamp/.display)
  • Use search.sh with project and service flags to narrow results and reduce noise

Example use cases

  • Bootstrap an agent with recent project prompts and summaries during /init
  • Answer a user question like “what did I change last week in this repo?” by searching session history
  • Extract user messages and bash commands to infer common workflows and preferred tools
  • Generate a short summary of recent sessions to onboard a new collaborator
  • Audit tool usage to see which functions or plugins were invoked most often

FAQ

The scripts automatically search history.jsonl, which retains global prompts even when per-project session files were cleaned up.

How do Claude and Codex file formats differ?

Claude uses timestamp in milliseconds and display fields (.timestamp/.display). Codex uses seconds and .text/.ts fields. The provided jq examples handle both.

Built by
VeilStrat
AI signals for GTM teams
© 2026 VeilStrat. All rights reserved.All systems operational
migrating-from-codex-and-claude-code skill by letta-ai/letta-code | VeilStrat