Memory & Context

List recent commits linked to agent sessions, filtered by branch or repo

Delete chosen agent memories after review and explicit confirmation

Resume the latest session in this directory and pick up open questions

Save corrections as weighted lessons that resurface before similar work

Recall past memory before tasks, save decisions and learn from corrections

Search past sessions and learnings with hybrid keyword, vector and graph recall

Summarize recent sessions in a project by date with key highlights

Save decisions and insights to long-term agent memory with searchable tags

Show a clean timeline of what happened in past sessions on this project

Give agents a four-tier memory that learns durable facts from past sessions

Cut token use by about 75 percent with terse replies that keep accuracy

Load and refresh company context for executive advisor skills with privacy rules

Log an approved board memo as durable company decision memory

Log board decisions in two memory layers and track overdue action items

Compact a conversation into a handoff document for a fresh agent

Design, review and prune agent memory so it keeps only what matters

Review auto-memory for stale entries and promotion candidates

Check auto-memory health with line counts, capacity and stale entries

Promote proven learnings from auto-memory into permanent project rules

Save an important discovery to auto-memory with timestamp and context

Curate auto-memory into durable project rules and reusable skills

Compact a session into a redacted handoff file that the next session loads

Consolidate past sessions overnight into memory and skill improvements

Persistent file-based planning for multi-step AI agent work and long-running tasks

Keep task plans, findings and progress files on disk across long agent sessions

Persistent file-based planning with task, findings and progress files for agent work

File-based planning that survives context resets through task and progress files

Persistent planning files for multi-step agent work that survive context compaction

Persistent file planning for multi-step agent tasks with recovery after context clears

Build your own profile and distill colleagues or public figures into reusable profiles

Record a non-obvious fix as a durable repo learning for future sessions

Audit stored repo learnings for staleness and conflicts with the current code

Write a session handoff so another agent can continue without chat history

Switch to ultra-compressed output to save tokens on large tasks

Read, write and curate the project memory folder that records what the project knows

Keep per-user memory in a private Drive folder that carries across all sessions

Search and update a typed knowledge archive of people, companies and research

Refresh personal goals and current state through an evidence-backed interview

Import notes and instructions from Obsidian, Notion and other tools into memory

Trim a bloated always-on context file safely with reversible, human-gated edits

Compress long agent sessions into handoff summaries that keep decisions and next steps

Diagnose context poisoning, clashes and lost-in-the-middle failures in agents

Get a collection of context engineering skills for multi-agent and production systems

Explain context window anatomy, attention limits and why context quality matters

Cut token cost with context budgeting, observation masking and KV-cache strategy

Keep agent context in files with scratchpads, tool output offloading and handoffs

Share memory between worker agents with KV cache compaction instead of summaries

Build persistent agent memory with entity tracking, temporal validity and graph retrieval

Let an agent edit its own live context window with eviction and in-place updates