ce-sessions

Search and synthesize coding agent session history across Claude Code, Codex, and Cursor.

Updated May 11, 2026
One-click install
npx skills add https://github.com/mmnavarr/harness --skill ce-sessions-mmnavarr
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ce-sessions
Source: https://github.com/mmnavarr/harness/tree/main/skills/ce-sessions
Command: npx skills add https://github.com/mmnavarr/harness --skill ce-sessions-mmnavarr

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? When you return to a codebase, prior work done in AI coding sessions is scattered across Claude Code, Codex, and Cursor transcript files that are too large to read manually. This Skill answers questions about what was tried, decided, or learned in past sessions without loading multi-megabyte JSONL files into context. ## Core Features & Use Cases - Cross-platform session discovery: Locates session files across Claude Code, Codex, and Cursor, filtered by repo, time window, branch, and keyword matches. - Token-safe extraction: Scripts extract metadata, conversation skeletons, and error signals to scratch files so bulk content never floods the orchestrator's context. - Subagent synthesis: Dispatches a historian subagent to read extracted skeletons and return structured findings (what was tried, what failed, key decisions). - Use Case: Ask "what did we try last week to fix the auth middleware token rejection bug?" and get a synthesized summary of relevant prior sessions across all three platforms. ## Quick Start Ask the agent to search your past coding sessions for what was previously attempted on the current bug using the ce-sessions skill.

Frequently Asked Questions about ce-sessions

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

FAQPage Schema
How do I search past Claude Code or Codex session history?▼

Invoke the skill with a question or topic, and it discovers session JSONL files across Claude Code, Codex, and Cursor, filters them by repo, branch, time window, and keywords, then synthesizes findings. Running it with no argument prompts you for what you want to know.

What platforms does session history search support?▼

The discovery script supports Claude Code, Codex, and Cursor. You can restrict the scan to one platform with the --platform flag, or search all three by default.

Can I filter session search by git branch or keyword?▼

Yes. Claude Code sessions are filtered by exact or topic-matching git branch, and a keyword fallback scans user and assistant message text across all platforms. Codex sessions skip branch filtering because they do not record gitBranch.

Why does session search return no relevant prior sessions?▼

Common causes are a scan window that is too narrow, sessions older than Claude Code's roughly 30-day default retention, or keyword matches of zero. The skill starts with a narrow window and widens only when nothing relevant is found.

Does searching session history expose sensitive content?▼

The skill never reads entire session files into context and never reproduces tool inputs, outputs, or thinking blocks verbatim. Guardrails direct the synthesizer to surface technical content only and exclude credentials or personal material.