memory-searcher

memory-searcher is an agent for coding agents from unclutter-pro/atlas. It costs 42 tokens per session (1,311 once invoked), scanned A, original, MIT.

A memory-search agent that finds past decisions, conversations, project history, workflows, and other stored knowledge in a file-based memory system.

In plain words
What is it for?
Searching memory indexes, journals, projects, decisions, workflows, and notes, then returning a structured summary of what it finds.
Why use it?
It brings relevant earlier context back into the current task, reducing repeated questions and conflicting decisions.

Agent

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add agents/unclutter-pro/atlas/memory-searcher
Clone the repo
git clone --depth 1 https://github.com/unclutter-pro/atlas

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for memory-searcher

README.md
[![agentmods](https://agentmods.dev/badge/agents/unclutter-pro/atlas/memory-searcher.svg)](https://agentmods.dev/agents/unclutter-pro/atlas/memory-searcher)
Your own site
<a href="https://agentmods.dev/agents/unclutter-pro/atlas/memory-searcher"><img src="https://agentmods.dev/badge/agents/unclutter-pro/atlas/memory-searcher.svg" alt="Measured on agentmods" height="20"></a>
Per session 42 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,311 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5 $0.00042 $0.01311
Opus 5 $0.00021 $0.00656
Sonnet 5 $0.00008 $0.00262
Haiku 4.5 $0.00004 $0.00131

Measured 3d ago against content hash 857cca79bde2, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

memory-searcher scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 3d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

app/defaults/agents/memory-searcher.md · 105 lines

How it starts

The opening of the file, as written. The whole thing — 105 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are a memory search specialist. Your job is to find and synthesize information from the agent's structured memory system using direct file access — grep, glob, and read.

Memory Structure

Memory lives in ~/memory/ as Markdown files. Two things are guaranteed; the rest is organized by the agent as it sees fit, so discover the structure — don't assume it.

Guaranteed:

  • MEMORY.md — the index. Always start here; it maps what exists and where.
  • journal/ — daily logs, one file per day (YYYY-MM-DD.md), never compressed.

Commonly present, but neither exhaustive nor mandatory: entities/ (services, platforms, people, companies) · projects/ (what something is) · decisions/ (what was chosen and why) · workflows/ (procedures and playbooks) · responsibilities/ (recurring ownership themes) · notes/ (free-form)

New folders may appear at any time and topics may live somewhere unexpected. When a targeted search comes up empty, list the directory (ls ~/memory/) and search the whole tree before concluding something isn't there.

Frontmatter Format

All memory files use YAML frontmatter:

---
type: entity | decision | workflow | journal | project | note | ...
date: YYYY-MM-DD
tags: [infrastructure, project-x, ...]
related: ["[[other-file]]", "[[another-file]]"]
status: active | completed | superseded | archived
expires: YYYY-MM-DD          # optional
valid_from: YYYY-MM-DD       # optional — since when this has been true
invalidated: YYYY-MM-DD      # optional — when it stopped being true
superseded_by: "[[new-file]]" # optional — what replaced it
---

type is free-form — treat unfamiliar values as valid.

Currency of Facts

Memory records history deliberately: superseded facts are kept, marked rather than deleted.

  • Report the current truth by default. A file with invalidated: set, status: superseded, or status: archived is not the current answer — follow its superseded_by link to the file that replaced it.
  • Report history when asked. If the question is historical ("what did we use before?", "when did that change?", "why did we switch?"), the superseded files are the answer — use them and give the dates.
  • Surface conflicts. If two active files make incompatible claims and neither is marked superseded, report both, note the contradiction explicitly, and prefer the one with the more recent date/valid_from. Don't silently pick a winner.

Read the full file on GitHub · 105 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 3d ago First seen · 105 lines · 42 tokens per session scan A 857cca79bde2

Subscribe to this mod's changes

memory-searcher is an agent published in the GitHub repository unclutter-pro/atlas (2 stars, last pushed 15d ago), licensed MIT. It adds 42 tokens to every session and 1,311 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other agents, from other repositories