cog CLAUDE.md

cog CLAUDE.md is an instructions file for coding agents from marciopuga/cog. It costs 1,595 tokens per session, scanned A, original, MIT.

Instructions for Cog, a persistent-memory system that stores plain-text files across sessions. They define its memory locations, three memory levels, persona rules, and project conventions.

In plain words
What is it for?
Use them when configuring or operating Cog memory. They cover the hot, warm, and archived memory areas, the COG_HOME location, and the required behavior for an agent using that memory.
Why use it?
They give an agent consistent rules for reading and writing long-term memory, including where files belong and how much information should be loaded routinely.

Instructions file

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 instructions/marciopuga/cog/claude-md
Clone the repo
git clone --depth 1 https://github.com/marciopuga/cog

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 cog CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/marciopuga/cog/claude-md.svg)](https://agentmods.dev/instructions/marciopuga/cog/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/marciopuga/cog/claude-md"><img src="https://agentmods.dev/badge/instructions/marciopuga/cog/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 1,595 This file is loaded in full into every session.
When invoked 1,595 The same file — it is already loaded in full.
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 $0.01595 $0.01595
Opus 5 $0.00797 $0.00797
Sonnet 5 $0.00319 $0.00319
Haiku 4.5 $0.00160 $0.00160

Measured 5d ago against content hash eb1c5c7eed0f, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

cog CLAUDE.md 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 5d 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.

CLAUDE.md · 125 lines

How it starts

The opening of the file, as written. The whole thing — 125 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Cog — Memory System

Cog gives you persistent memory across sessions. Memory lives in memory/ as plain text files.

Core Conventions

The full memory conventions are defined in the cog skill (.claude/commands/cog.md for Claude Code — vendored from cog-skills, which is the canonical source). The key rules are summarized below; when in doubt, the cog skill wins.

Memory Path

All memory paths resolve against $COG_HOME/memory/ if the COG_HOME env var is set, otherwise ~/cog/memory/. One folder, many projects — never scaffold memory inside individual projects.

Persona

  • Think and speak as an extension of your owner — their values, their voice, their priorities
  • Concise, proactive, direct — no filler
  • When uncertain, say so plainly
  • Write immediately — don't wait to save something worth remembering

Memory System

Persistent memory lives in memory/. Three tiers:

  • Hot (memory/hot-memory.md — the root file only) — loaded every conversation, <50 lines, rewrite freely
  • Warm (domain files, including each domain's own hot-memory.md) — loaded when the domain skill activates
  • Glacier (memory/glacier/) — YAML-frontmattered archives, indexed via glacier/index.md. Read-only, except housekeeping archival.

L0 Headers

Every memory file has <!-- L0: summary (max 80 chars) --> as line 1 — including auto-generated indexes.

Memory Retrieval Protocol

When responding to any query:

  1. Identify domain — match query to a domain (see memory/domains.yml)
  2. L0 scangrep -rn "<!-- L0:" memory/{domain}/ to find relevant files (memory/{domain}/INDEX.md is a precomputed table of the same headers)
  3. Select by query type:
    • Tasks → action-items.md + calendar.md
    • Person → entities.md
    • Overview → hot-memory.md + action-items.md
    • Cross-reference → check link-index.md
  4. L1 before L2 — for files >80 lines, scan section headers before full read
  5. SSOT check on write — before writing, verify the fact doesn't already exist elsewhere

Read the full file on GitHub · 125 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. 5d ago First seen · 125 lines · 1,595 tokens per session scan A eb1c5c7eed0f

Subscribe to this mod's changes

cog CLAUDE.md is an instructions file published in the GitHub repository marciopuga/cog (375 stars, last pushed 2mo ago), licensed MIT. It adds 1,595 tokens to every session, about $0.0080 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 2026-08-30.