memory

Two-layer memory system with grep-based recall.

Install with Tessl CLI

npx tessl i github:HKUDS/nanobot --skill memory

What are skills?

1.30x

Quality

47%

Does it follow best practices?

Impact

95%

1.30x

Average score across 3 eval scenarios

Optimize this skill with Tessl

npx tessl skill review --optimize ./nanobot/skills/memory/SKILL.md

SKILL.md

Review

Evals

Evaluation results

100%

25%

Personal Assistant Memory Bootstrap

Two-file memory structure

Criteria

Without context

With context

MEMORY.md location

100%

HISTORY.md location

100%

History entry timestamp format

100%

User preferences in MEMORY.md

100%

Project context in MEMORY.md

100%

Relationships in MEMORY.md

100%

Events in HISTORY.md

100%

HISTORY.md append-only pattern

100%

MEMORY.md written via file tool

100%

Files in separate roles

100%

Without context: $0.4178 · 1m 50s · 20 turns · 25 in / 6,299 out tokens

With context: $0.3215 · 1m 17s · 19 turns · 57 in / 4,476 out tokens

90%

26%

AI Assistant History Recall Module

History search method selection

Criteria

Without context

With context

Correct history file path

100%

Small-file read approach

100%

Large-file exec approach

100%

Size-based branching

100%

grep command for Linux/macOS

40%

100%

Cross-platform alternative

Case-insensitive search

100%

README describes both strategies

100%

Returns matching lines only

100%

Without context: $0.2326 · 1m 4s · 14 turns · 20 in / 3,379 out tokens

With context: $0.3317 · 1m 6s · 22 turns · 427 in / 3,536 out tokens

96%

16%

Migrate Developer Context into Persistent Memory

Fact categorization and immediate storage

Criteria

Without context

With context

MEMORY.md path

100%

HISTORY.md path

100%

Preferences in MEMORY.md

100%

Project context in MEMORY.md

100%

Relationships in MEMORY.md

100%

Events in HISTORY.md not MEMORY.md

100%

HISTORY.md entry format

100%

MEMORY.md written via file operation

100%

HISTORY.md append semantics

20%

Stable facts not in HISTORY.md

100%

Without context: $0.3055 · 1m 17s · 17 turns · 24 in / 4,466 out tokens

With context: $0.4891 · 1m 55s · 26 turns · 283 in / 6,560 out tokens

Evaluated: 1 day ago
Agent: Claude Code
Model: Claude Sonnet 4.6

Table of Contents

Personal Assistant Memory Bootstrap AI Assistant History Recall Module Migrate Developer Context into Persistent Memory

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.