Cross chat todo persistence
Skill jpoindexter/awwwards-todo-skills/skills/cross-chat-todo-persistence
Use when todo/task sessions must survive across separate conversations or agent runs — defines the session data model, save/load/switch operations, incomplete-session detection, and the new-conversation resume flow so an agent can pick up unfinished work from a prior chat.From its SKILL.md
npx -y skills add jpoindexter/awwwards-todo-skills --skill cross-chat-todo-persistenceAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
2 things to look at
- 29 days oldThe repository was created 29 days ago. New is not bad, but a brand new repository carrying a familiar-sounding name is the shape a typosquat arrives in, and there has been no time for anyone else to find a problem with it.
- 0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
SKILL.md
3.0 KB, 700 tokens by cl100k_base, as published. Nobody here has run it
Cross-Chat Todo Persistence
Keep todo sessions alive between conversations. On a new chat, detect unfinished work, resume the most recent one, and offer to switch. This is the state concern only — render with swiss-grid-todo-display, decide when to surface with contextual-todo-triggers.
When to use
- Work spans multiple sessions and the user expects continuity ("where were we?").
- You need to manage several parallel projects and switch focus between them.
Session data model
{
"id": "design-system-project",
"title": "Design System Project",
"todos": [
{ "content": "Review MCP server documentation", "status": "completed", "priority": "high" },
{ "content": "Test Swiss Grid implementation", "status": "in_progress", "priority": "medium" }
],
"lastModified": "2026-07-18T10:00:00.000Z",
"lastWorkedOn": "2026-07-18T10:00:00.000Z"
}
status ∈ pending | in_progress | completed | cancelled. lastWorkedOn drives recency ordering.
Operations
| Operation | Behavior |
|---|---|
save(session) | Upsert by id; stamp lastModified/lastWorkedOn. |
loadAll() | Return every stored session keyed by id. |
getIncomplete() | Sessions with at least one pending or in_progress item. |
getMostRecentIncomplete() | Incomplete sessions sorted by lastWorkedOn desc → first. |
switch(id) | Set current session, re-stamp lastWorkedOn, return it. Fail clearly if id missing. |
New-conversation resume flow
- On chat start, call
getIncomplete(). - Zero incomplete → "No active todos found. Ready to create a new session!"
- One or more → resume
getMostRecentIncomplete()as current, render it, then list the others as an overview:• {title} — {completed}/{total} complete, {inProgress} in progress ({lastWorkedOn date})
- Ask once: "Continue with {title}, or switch to a different session?"
Storage
Runtime-agnostic default: a single JSON file, one object keyed by session id.
Claude Desktop-specific (isolated): the original MCP server stored at
~/Library/Application Support/Claude/todos.json. On any other runtime, use a path you control (e.g.$XDG_DATA_HOME/todos.jsonor./.todos.json). The reference implementation takes the path as a constructor argument — no host assumption baked in.
Reference implementation
reference/session-store.mjs — self-contained, file-backed ES module. Constructor takes a storage path; exports the operations above. Run node reference/session-store.mjs for a demo against a temp file.
Agent-runtime notes
The data model and flow are runtime-agnostic. Only the default storage path is host-specific and is isolated in the reference constructor argument, not hard-coded.
What ships with it: 1 file
3.8 KB alongside SKILL.md, 1 of them executable
reference/
- session-store.mjsruns3.8 KB