- Home
- Skills
- Productivity
- resume-memo: save session state before /compact or /clear
Works with the AI Agents you already use
resume-memo: save session state before /compact or /clear
Writes the state that exists only in your conversation (decisions and why, failed attempts, observations kept apart from inferences, next steps) to a YAML memo, so a fresh session can continue from the file alone.
Free
resume-memo: save session state before /compact or /clear
Example session with this skill installed
Save a resume memo before I run /clear.
- Read your context and instructions
- Compiled the resume-memo
The memo below is an illustrative sample for a CSV export bug, not a captured session.
task: fix csv export
created: 2026-11-13 10:00
status: open
cwd: /work/myapp
branch: fix/csv-export
head: 1a2b3c4
goal: Fix the CSV export that drops the last row
decisions:
- what: stream the rows instead of building the whole file in memory
why: the failing export is 2 GB
rejected: raising the memory limit
observations:
- fact: export of 10 rows returns 9 rows
inference: off-by-one in the batch loop
basis: unverified
next:
- add a failing test for the 10-row case
- fix the batch loop in app/exports/csv.rb:42
The skill ends by telling you where the memo is and what to paste in the next session:
Memo saved: /Users/you/.claude/resume-memos/20261113-1000-myapp-fix-csv-export.yml
After /clear, paste this as your first message:
Read /Users/you/.claude/resume-memos/20261113-1000-myapp-fix-csv-export.yml and resume the work
The inference "off-by-one in the batch loop" rests on one observation, so its basis is unverified. When the memo is restored, the skill tells Claude not to copy an unverified cause into an Issue, a PR or a spec.
Connects securely to your tools. The creator never sees your data.
About this skill
/compact summarizes a conversation and /clear discards it. How a decision was reached, what was investigated and what failed can be lost or distorted either way. This skill saves that state to one YAML file per session before you run either command, and a fresh session that reads the file can continue the work.
What goes into a memo
- Decisions and the reasons for them ("why B and not A"), and approaches that were tried and failed.
- Observations (commands run, their output) as separate items from the inferences drawn from them. Each inference carries a
basis, and an inference from a single observation is markedunverified. - Open questions, the user's standing instructions, and numbered next steps.
- Not file contents, diffs or secrets. Those can be recovered from disk or git, or should not be written at all.
How it works
- Say "save the session state before I compact", or announce
/compactor/clear. The skill does not run when you are only talking about compaction. - It writes
~/.claude/resume-memos/<YYYYMMDD-HHMM>-<project>-<task-slug>.yml. SetRESUME_MEMO_DIRto use another directory. - It tells you the full path and a sentence to paste into the next session:
Read <full path> and resume the work. - In the new session, Claude reads the memo as data recorded earlier, not as instructions, and compares the recorded
branchandheadwith the repository before relying on it.
What this package does not include
This package contains only SKILL.md. Two parts of the project are in the plugin on GitHub, not here:
- The check script that runs after the memo is written. It reports list items that YAML read as something other than text, and secret-like text.
SKILL.mdcalls it asscripts/check-memo.sh; without the plugin that step fails because the script is missing. - The hooks. After a compaction,
SessionStartprints the head of the newest open memos of the current project, andPreCompactprints instructions for the summary.
Requirements: git (the memo records the branch and HEAD) and bash. The check script in the plugin needs ruby, or python3 with PyYAML.
To get the check script and the hooks, install the plugin:
/plugin marketplace add sai-kaneko-31/cc-resume-memo
/plugin install resume-memo@cc-resume-memo
Source: https://github.com/sai-kaneko-31/cc-resume-memo (MIT)
日本語の説明
resume-memo は、会話の中にしか無い作業の状態を、1 セッション 1 ファイルの YAML メモに書き出すスキルです。/compact は会話を要約し、/clear は会話を捨てます。どちらを実行しても、判断の理由、調べて分かったこと、失敗した試みが、消えたりゆがんだりすることがあります。/compact か /clear を実行する前に、このスキルがそれらをメモへ残します。新しいセッションは、そのメモだけを読んで作業を続けられます。観察(実行したコマンドと出力)と、そこから導いた推測は別の項目に分けて、推測には根拠(basis)を付けます。1 回の観察だけから導いた推測は、basis に unverified と書きます。ディスクや git から取り戻せるもの(作業中のファイルの中身、diff)と、秘密の値は、メモに書きません。メモを書いたあとの YAML の確認と、compact の前後で動く hook は、GitHub(https://github.com/sai-kaneko-31/cc-resume-memo)の plugin 版にだけ入っています。この ZIP は SKILL.md だけなので、その 2 つは動きません。
Writes the memo to ~/.claude/resume-memos/ (or the directory in RESUME_MEMO_DIR), so it needs a local filesystem. This package contains only SKILL.md: the check script and the hooks come with the plugin on GitHub. The check script needs ruby, or python3 with PyYAML. Native Windows is not supported (WSL should work; it has not been tried).
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 any payment.
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
Free forever
No account required to browse
Trust & safety
Security scan
Security scan passed 4 days ago
- Free with an Agensi account
Requires
Permissions
File Scopes