- Home
- Skills
- Agents & Orchestration
- The Archivist: keep CLAUDE.md lean, fresh and cheap
More screenshots
Works with the AI Agents you already use
The Archivist: keep CLAUDE.md lean, fresh and cheap
Your CLAUDE.md only grows. This dates every line, re-checks the old ones, moves the dead ones to an archive and tells you what each session costs.
$9
The Archivist: keep CLAUDE.md lean, fresh and cheap
Example session with this skill installed
Our CLAUDE.md has grown messy and a lot of it is out of date. Some of it is copied from docs/testing.md. Please clean it up.
- Read your context and instructions
- Compiled the the archivist
CLAUDE.md: 510 -> 264 words, ~663 -> ~343 tokens per session. Stale lines 4 -> 1, unmarked 32 -> 0, dead paths 6 -> 0.
Actions: 20 marked, 1 refreshed, 17 archived (moved, nothing deleted)
- refreshed CLAUDE.md:8 "Orders are handled in src/api/orders.ts" evidence: ls src/api shows orders.ts
- archived CLAUDE.md:30 "Code freeze for the holiday release." reason: until=2026-11-10 has passed
Left for you: CLAUDE.md:22 Staging DB is read-only. Check: ask ops if staging is writable again.
Connects securely to your tools. The creator never sees your data.
About this skill
Forgets on purpose.
Agent instruction files like CLAUDE.md, AGENTS.md, GEMINI.md and .cursorrules only grow. Paths get renamed, a code freeze ends, a rule gets copied from the testing doc, and every session still loads all of it.
The Archivist gives each line a short dated marker, hidden in Markdown as an HTML comment, with one of three tiers:
- pinned: team rules that never expire
- aging: facts about the code, stale after 30 days
- perishable: temporary state, stale after 7 days, and it must say what to check
How it works:
scanlists stale, expired and unmarked lines withfile:line. It also flags backticked paths that no longer exist, long command blocks, history lines, and words and tokens per file against a budget.- The agent re-checks each stale line against the repo today. It either refreshes the date with evidence it can name, or moves the line to
CLAUDE.archive.mdwith a dated reason. Nothing is deleted. dupsfinds near-duplicate rules across your agent file and your docs. One place owns each rule, and the others get a one-line pointer.- Long commands and history move out of the every-session file into docs,
--helpor git. reportshows a before and after table and every refresh and archive with its reason.
The agent asks before archiving pinned lines, moving text into docs, or editing lines the scan did not report.
How it differs from conflict and staleness resolvers: those audit memory logs for contradictions and output a context pack. This one keeps your live instruction files in the repo small, with a date on each line, an archive you can restore from, and a rule for where each kind of guidance should live.
For teams and solo developers whose agent file is past 100 lines.
Part of The Guild, agent skills by Sesisnando Neto: The Oathkeeper (sealed acceptance checks), The Quartermaster (quests and decisions), The Tracker (root-cause proof), The Lorekeeper (repo principles from history), The Inspector (QA validation).
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Agents & Orchestration.
Token counts are estimates (words x 1.3), not a real tokenizer. Duplicate detection compares words, not meaning, so the same rule in very different words is missed. The missing-path check only looks at backticked paths, so placeholder paths in examples can show up. archive moves lines and their child bullets, not code blocks or tables; the agent moves those with an edit and reports it. Lines about people or outside systems cannot be checked from the repo and are left for the human.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed today
- Software
Python 3.8+
- Optional
Git (finds the repo root)
Permissions
File Scopes
No network. Reads the memory files you name, writes <name>.archive.md next to each and a small log in .memory-decay/. Never deletes a line; archive is a move.