How do I install the Publish Session skill?

Run npx skillmds@latest add labspangaea/publish-session in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Publish Session safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Publish Session?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Publish Session free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Publish Session?

labspangaea (@labspangaea) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Publish Session by labspangaea — Agent Skill

Name: Publish Session
Author: labspangaea

Publish Session

Hand the CURRENT conversation off to another Claude Code profile / subscription by copying this session's transcript into a shared staging directory (~/claude-profiles/shared-conversation/) that every profile on the machine can read. Use this whenever the user wants to continue, resume, move, hand over, or share what you are working on right now with their other account — phrasings like "share this session with my work profile", "I want to keep going on this under claude-work", "hand this conversation over to my personal subscription", "export this session so my other login can pick it up", "publish this conversation", "I'm running out of limits, move this to the other account", or "continue this in claude-personal". Also trigger when the user hits a rate limit or plan cap mid-task and asks how to carry the context to their other subscription. Each profile sets its own CLAUDE_CONFIG_DIR and therefore has a completely separate session list, so a session recorded under one profile is invisible to `claude --resu

labspangaea faa6f22 2 files · 7.0 KB Updated Sep 17, 2026

File contents

Publish this session to the shared hand-off directory

A Claude Code profile is just a CLAUDE_CONFIG_DIR. Transcripts live under $CLAUDE_CONFIG_DIR/projects/<slugified-cwd>/<session-id>.jsonl, so a session recorded by the default ~/.claude profile simply does not exist as far as a ~/.claude-work profile is concerned. Publishing copies the transcript somewhere both can see.

PLUGIN_DIR below is this skill's own directory — the publish-session/ folder holding this file, whether that is inside the installed plugin or a standalone ~/.agents/skills/publish-session.

Step 1 — stamp a nonce into the transcript

Run this on its own, in its own Bash call, with a fresh random string each time:

echo "ptl-<8 random hex chars>"

Step 2 — publish, in a LATER Bash call

python3 PLUGIN_DIR/scripts/publish_session.py --nonce ptl-<the same string>

The two calls have to be separate. The transcript is flushed only once a tool call returns, so a nonce echoed in the same command the script runs in will not be on disk yet and the script will exit 1 telling you so — if that happens, just echo a new nonce and try again.

Why a nonce at all: the obvious "newest .jsonl in this project directory" heuristic picks the wrong file whenever the user has two sessions open in the same project, which is exactly the situation someone juggling two subscriptions is in. Grepping for a string only this conversation contains is unambiguous.

Step 3 — tell the user how to pick it up

The script prints the session id and the published path. Relay both, plus the command for the other profile:

/claude-profiles:consume-session <session-id>

Things worth saying out loud

The copy is a snapshot. The live transcript keeps growing as this conversation continues, so anything said after publishing is not in the copy. If the user wants the last few turns included, publish again at the end — re-running just overwrites the same file.
The transcript is the whole conversation, including file contents and command output that scrolled past. If this session touched credentials or client data, say so before publishing — the staging directory is plain files in the user's home, readable by anything running as them.
Nothing is deleted or moved. The original session stays exactly where it is and remains resumable under this profile.

---name: publish-sessiondescription: Hand the CURRENT conversation off to another Claude Code profile / subscription by copying this session's transcript into a shared staging directory (~/claude-profiles/shared-conversation/) that every profile on the machine can read. Use this whenever the user wants to continue, resume, move, hand over, or share what you are working on right now with their other account — phrasings like "share this session with my work profile", "I want to keep going on this under claude-work", "hand this conversation over to my personal subscription", "export this session so my other login can pick it up", "publish this conversation", "I'm running out of limits, move this to the other account", or "continue this in claude-personal". Also trigger when the user hits a rate limit or plan cap mid-task and asks how to carry the context to their other subscription. Each profile sets its own CLAUDE_CONFIG_DIR and therefore has a completely separate session list, so a session recorded under one profile is invisible to `claude --resu---# Publish this session to the shared hand-off directoryA Claude Code profile is just a `CLAUDE_CONFIG_DIR`. Transcripts live under`$CLAUDE_CONFIG_DIR/projects/<slugified-cwd>/<session-id>.jsonl`, so a session recorded by thedefault `~/.claude` profile simply does not exist as far as a `~/.claude-work` profile is concerned.Publishing copies the transcript somewhere both can see.`PLUGIN_DIR` below is this skill's own directory — the `publish-session/` folder holding this file,whether that is inside the installed plugin or a standalone `~/.agents/skills/publish-session`.## Step 1 — stamp a nonce into the transcriptRun this **on its own**, in its own Bash call, with a fresh random string each time:```bashecho "ptl-<8 random hex chars>"```## Step 2 — publish, in a LATER Bash call```bashpython3 PLUGIN_DIR/scripts/publish_session.py --nonce ptl-<the same string>```The two calls have to be separate. The transcript is flushed only once a tool call returns, so anonce echoed in the same command the script runs in will not be on disk yet and the script willexit 1 telling you so — if that happens, just echo a new nonce and try again.**Why a nonce at all:** the obvious "newest `.jsonl` in this project directory" heuristic picks thewrong file whenever the user has two sessions open in the same project, which is exactly thesituation someone juggling two subscriptions is in. Grepping for a string only this conversationcontains is unambiguous.## Step 3 — tell the user how to pick it upThe script prints the session id and the published path. Relay both, plus the command for theother profile:```/claude-profiles:consume-session <session-id>```## Things worth saying out loud- **The copy is a snapshot.** The live transcript keeps growing as this conversation continues, so anything said after publishing is not in the copy. If the user wants the last few turns included, publish again at the end — re-running just overwrites the same file.- **The transcript is the whole conversation**, including file contents and command output that scrolled past. If this session touched credentials or client data, say so before publishing — the staging directory is plain files in the user's home, readable by anything running as them.- **Nothing is deleted or moved.** The original session stays exactly where it is and remains resumable under this profile.

Frequently asked questions