emihiggins

Agents Recap — AI skill for Claude Code

AI community

Local, offline recap and RAG memory of your AI coding sessions.

How to install Agents Recap

This entry records only its repository, not the path inside it, so there is no exact command to give. Open emihiggins/agents-recap and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Agents Recap does

Local, offline recap and RAG memory of your AI coding sessions. Reads Claude Code, Cursor and VS Code Copilot Chat session state read-only, tells you where you left off, and lets you ask about it later.

Alternatives in AI

Academic Humanizer — Strip AI-writing tells from papers and grant proposals (NSF/NIH), while keeping scholarly voice and tying clai 1.1k ★
Late CLI — High-performance AI agent for long-horizon tasks and local models 442 ★
Cross Agent Session Resumer — Resume AI coding sessions across providers: converts Codex, Claude, Gemini, and other session formats through 71 ★

README

agent-recap

You had six AI sessions going. Which one still has work in it?

`agent-recap` reads the session state your AI tools already keep on disk, works out where each project actually stands, and writes one HTML page telling you where you left off. It also remembers that context so you can ask about it later.

Everything runs locally against [Ollama](https://ollama.com). No API keys, no network calls, nothing leaves the machine.

agent-recap                       # build the recap and open it
agent-recap ask "what was I doing with the rates API?"
agent-recap status                # what's running right now
agent-recap schedule --at 08:30   # have it ready before you sit down

![The recap page: a waiting-on-you banner, then one card per project with its recap, suggested next step, git state and unfinished work](docs/recap.png)

*Sample data. Regenerate with `python scripts/sample_recap.py`.*

It leads with what's actually blocking you:

Waiting on you
  storefront-web — permission prompt, waiting 48m

An agent stalled on a permission prompt for the last three quarters of an hour is the single most useful thing this can tell you, so it gets a banner.

What it reads

Source	Where	What it gets
Claude Code	`~/.claude`	Live session status, titles, last messages, plan documents and their remaining steps
Cursor	Cursor's global `state.vscdb`	Composer/agent sessions, titles, agent todo lists, context usage
VS Code Copilot Chat	VS Code `workspaceStorage`	Chat sessions (both the legacy and journal formats), Copilot todo lists
git	each project directory	Branch, uncommitted files, unpushed commits

All of it is opened **read-only**. Cursor and VS Code are usually running while this executes, so connections are opened with `mode=ro` and the test suite asserts that writes and DDL are rejected. `agent-recap` never modifies another tool's data.