What sources does recall search?

Recall searches specific memory files: user.md, soul.md, work.md, skills index and project sessions; for broad topics it can grep across ~/.katana/memory.

How are results presented?

Results are shown as a concise list with date and project tag, preceded by a header like Found X entries for 'term'.

How do I search broadly if I have no specific term?

Use full text grep across memory folder to surface any mention of the topic.

recall

Scanned

npx machina-cli add skill shane9coy/katana-agent/recall --openclaw

Files (1)

SKILL.md

1.7 KB

Recall — Retrieve Memories

When This Fires

User types /recall with a query
User asks: "what did we work on yesterday?"
User asks: "what do you know about me?"
User asks: "what's the status of project X?"
User asks: "do you remember when we..."
User references a past session or project

How To Search

Determine what user is asking for

Question type	Search where
"What do you know about me?"	`~/.katana/memory/core/user.md`
"How should you behave?"	`~/.katana/memory/core/soul.md`
"What did we work on?" / project status	`~/.katana/memory/work.md`
"What skills do we have?"	`~/.katana/memory/skills/_index.md`
Unclear	Search all files

Search work.md

Entries are formatted as:

## YYYY-MM-DD — project-tag
Summary of work done...

Filter by project tag, date, or keyword.

Search projects/

Check ~/.katana/memory/projects/{name}/sessions.md for project-specific history.

Full-text search

For broad queries, use grep across all .md files:

grep -r -i "search term" ~/.katana/memory/ --include="*.md"

Response Format

Present findings concisely:

🔍 Found 3 entries for "redis":

1. 2026-02-19 — redis-cache
   Implemented connection pooling. Fixed deadlock in transaction handler.

2. 2026-02-15 — redis-cache
   Initial setup with Sentinel failover. Chose ioredis over node-redis.

3. 2026-02-10 — api-gateway
   Added Redis caching layer for auth tokens. 15ms → 2ms latency improvement.

Source

git clone https://github.com/shane9coy/katana-agent/blob/main/templates/universal/skills/recall/SKILL.mdView on GitHub

Overview

Recall searches and retrieves memories from past sessions, projects, and memory files. It is triggered by /recall or natural prompts like what did we work on or what do you know about me. It helps maintain continuity and context across conversations.

How This Skill Works

Recall maps your query to memory sources such as user.md, soul.md, work.md, and per project sessions. It can perform targeted lookups or full text searches and then returns concise results with dates and tags.

When to Use It

You want the status or outcomes of a past project
You ask what we worked on yesterday or last session
You ask what the agent knows about you
You reference a past session or project and need context
You need to search memory files for broad facts about skills or projects

Quick Start

Step 1: Type /recall followed by your query (e.g., /recall project status)
Step 2: Read the concise results returned with dates and tags
Step 3: Reference or cite the memory in your current task or decision

Best Practices

Use precise keywords that map to a specific memory source (e.g., project tag, date, or keyword)
Start with a targeted query to avoid unrelated results
Include a project name or date to narrow results
Validate retrieved items against dates and tags
When uncertain, run a full-text grep to surface broader matches

Example Use Cases

Found 2 memories for 'Apollo integration': 2026-03-01 — Apollo-integration: Implemented CMS integration; 2026-02-20 — Apollo-integration: Deployed feature flags
Found 1 memory for 'what we worked on yesterday': 2026-03-07 — api-gateway: Refactored auth flow
Found 3 memories for 'project X': 2026-02-15 — X: Initial plan; 2026-02-28 — X: Deployment script; 2026-03-03 — X: QA results
Found 2 memories for 'you know me': 2026-01-10 — user prefs: adjust tone; 2026-03-01 — memory core: updated soul profile
Found 1 memory for 'sessions.md': 2026-03-04 — session 42: sprint planning notes

Frequently Asked Questions

Add this skill to your agents