recall

Reconstruct recent working context from chat transcripts and shared records into a current-state brief.

Updated Mar 27, 2026
One-click install
npx skills add https://github.com/gmackie/agent-skills --skill recall-gmackie
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: recall
Source: https://github.com/gmackie/agent-skills/tree/main/skills/recall
Command: npx skills add https://github.com/gmackie/agent-skills --skill recall-gmackie

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? When resuming work after days away, you lose track of what you decided, which PRs shipped, and what problems remain. This Skill rebuilds your recent working context from your own chat history and the shared engineering record, then hands back a tight brief of where things stand and what to do next. ## Core Features & Use Cases - Chat History Mining: Fans out parallel subagents across Cursor agent transcripts to extract topics, decisions, open threads, and artifacts, each cited by chat UUID. - Shared Record Sweep: Searches source control, issue trackers, chat channels, docs, and error tracking for user-reported symptoms, shipped-and-reverted fixes, and production errors tied to a named feature or bug. - Live State Verification: Cross-checks surfaced PRs, branches, and tickets with git and gh so the brief reflects current truth, not stale history. - Use Case: Before resuming a half-finished feature, ask for a recap and receive a capsule summary, per-thread status tags like [open PR #N], the top recurring problems, and the single most useful next action. ## Quick Start Ask the agent to recall your work on the billing retry feature from the past two weeks and produce a current-state brief.

Frequently Asked Questions about recall

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

FAQPage Schema
How do I catch up on what I was working on in Cursor?▼

Ask the agent to recall your work on a topic, and it mines your recent chat transcripts, checks the shared record, and returns a brief with thread statuses, recurring problems, and the next move. The default window is the last 7 days.

How does chat transcript search work across multiple sessions?▼

Parallel subagents each take a slice of the transcript corpus, order candidates by real modification time, grep the topic first, and read only matching regions. The main thread receives only their structured findings, keeping context usage low.

Does recall work outside the Cursor editor?▼

The transcript mining relies on Cursor's agent-transcripts directory under ~/.cursor/projects, so it is designed for Cursor workspaces. The shared-record sweep uses git, gh, and issue trackers, which work in any repository.

When should I use session pickup instead of full recall?▼

Use session pickup when you want to resume one specific prior chat. Recall is for rebuilding working context across many recent chats and the shared record before starting or resuming broader work.

Why does recall check git and gh after reading transcripts?▼

Transcripts and stale tickets are history, not current truth. Verifying surfaced PRs, branches, and tickets against live state ensures the brief reflects what actually shipped, reverted, or remains open.