agentleFS
Sign inStart free
jin2004-cmdjin2004-cmd/codex-practical-skills/session-checkpoint/SKILL.md

session-checkpoint

Skill1 starsChanged 36 days agoMITView on GitHub

Maintain a compact continuation checkpoint for a long-running assistant project when its configured turn counter is due, a substantial task ends, or the user asks to prepare continuation. Not for unrelated chats or ordinary answers. It does not write or reorganize the memory base, which is personal-secretary-memory.

7 other projects use a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

7 rules. 0 are common, 5 are distinctive, and 2 are very specific to this project.

  • 0

    Common

    Most of the 6 edited versions we read keep these.

  • 5

    Distinctive

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use the project's configured counter for substantive user turns.

    line 13
  2. Distinctive:

    Run at the configured interval, a meaningful milestone, a workstream switch, or an explicit request.

    line 14
  3. Distinctive:

    If no run evidence exists, do not claim that automatic maintenance occurred.

    line 15
  4. Very specific:

    Read the current checkpoint and only the task-relevant memory entry.

    line 19
  5. Distinctive:

    Keep user statements, tool checks, suggestions, and unverified claims separate.

    line 21
  6. Very specific:

    Use scripts/session_checkpoint.py:

    line 26
  7. Distinctive:

    No chat history, no permissions change, no claim that maintenance ran automatically.

    line 48

“Common” is measured: kept by at least 6 in 10 of the 6 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Session Checkpoint

Maintain a small, factual disk checkpoint for a project that has a local AGENTS.md and a canonical memory area. This does not delete chat, change permissions, archive sessions, or trigger platform context compression.

When to use

  • Use the project's configured counter for substantive user turns; do not count tool updates, subagent messages, or no-information "continue" messages.
  • Run at the configured interval, a meaningful milestone, a workstream switch, or an explicit request.
  • If no run evidence exists, do not claim that automatic maintenance occurred.

Workflow

  1. Read the current checkpoint and only the task-relevant memory entry. Use the checkpoint format in references/checkpoint-format.md; do not reread the entire archive.
  2. Record the current goal, user-confirmed facts, evidence-backed outputs, authorization, unresolved blockers, and 1–3 next actions.
  3. Keep user statements, tool checks, suggestions, and unverified claims separate. Do not store secrets, company material, private relationship transcripts, or hidden chain-of-thought.
  4. Validate the checkpoint independently, then reset the counter only after the file is updated. Never clear a new turn with an old checkpoint.

Script

Use scripts/session_checkpoint.py:

python scripts/session_checkpoint.py --state <local-state.json> tick --turn-key <stable-key>
python scripts/session_checkpoint.py --state <local-state.json> status
python scripts/session_checkpoint.py --state <local-state.json> complete --checkpoint <checkpoint.md>

The state is local machine metadata. Do not copy it into a public memory repository.

Example

The checkpoint file stays small and factual:

# checkpoint 2026-10-07

当前任务：整理技能仓库 README
已完成：13 个 skill 的 frontmatter 补 version
未决：Remotion 两个 skill 是否拆得更细
下一步：给没有示例的 skill 补 Example
证据：git log 89f3cd0

No chat history, no permissions change, no claim that maintenance ran automatically.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

Reviews can't be read right now.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

Check your setup, free