cobusgreyling

Memtag — Documentation skill for Claude Code

Documentation community

Markdown memory hygiene for agent wikis, trust decay, context packing, and archive gc.

How to install Memtag

This entry records only its repository, not the path inside it, so there is no exact command to give. Open cobusgreyling/memtag and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Memtag does

Markdown memory hygiene for agent wikis, trust decay, context packing, and archive gc.

Alternatives in Documentation

Apply Self Review — Apply the approved improvements from the latest memory-self-review PROPOSAL 297 ★
Cog — 240+ Cognitive architecture for Claude Code -- persistent memory, self-reflection, and foresight via plain-tex 294 ★
Context Forge — CLI tool that scaffolds context engineering documentation for Claude Code projects 137 ★

README

memtag

[![CI](https://github.com/cobusgreyling/memtag/actions/workflows/ci.yml/badge.svg)](https://github.com/cobusgreyling/memtag/actions/workflows/ci.yml) [![PyPI](https://img.shields.io/pypi/v/memtag)](https://pypi.org/project/memtag/) [![License](https://img.shields.io/badge/License-Apache_2.0-blue.svg)](LICENSE)

**Keeps agent-written wikis trustworthy by tracking how memories decay, not just where they're stored.**

The next bottleneck in agent memory isn't storage or retrieval. It's **decay**. Notes rot: guesses expire, loops contradict each other, agent speculation gets laundered into fact. Retrieval tools make rot easier to find. memtag makes it visible and disposable.

memtag is a hygiene CLI + markdown convention for agent wikis: sense rot, select trusted context, sweep the vault.

Not an agent. Not a store.

memtag does **not** replace Grok, Cursor, Claude Code, or recollect.

It is a composable hygiene layer — pipe it between your vault and whatever CLI you run:

memtag pack ./vault --task "deploy API" --budget 8000   # select — the product
memtag lint ./vault --strict                            # sense — the sensor
memtag gc ./vault                                       # sweep — the janitor

Pipe `pack` output into your agent harness. Store long-term semantic memory in **recollect**. Govern markdown wikis with **memtag**.

recollect search "deploy API" | memtag pack ./vault --stdin --task "deploy API" --budget 8000

Why memtag

**Declared trust is a prior. Derived trust is the product.**

Everyone else asks the agent "how confident are you?" and stores the number. That's circular — you're trusting an agent's self-report to decide which agent-writing to trust.

memtag treats declared `confidence` as a starting guess, then computes `trust` from behavior: provenance, expiry, supersedes chains, contradictions, human touch. **Trust isn't a field the agent writes. It's a score memtag computes.** `pack` and `gc` rank on deriv