board-plan
Skill1 starsChanged 13 days agoMITView on GitHub
Break a task into items in this session's plan file, route each to a role the active mode allows, and keep the plan in step with the work (status, commits, decisions). Use whenever work splits into 2+ pieces or starts an agent.
In this repository5 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- configSkill
- software-standardsSkill
- test-inventorySkill
- working-modeSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈2
Distinctive · estimate
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
- Distinctive: line 10
When several sessions and agents work at once — across more than one project — the chat alone no longer shows who is doing what, what is …
- Distinctive: line 27
Create it before the first piece starts, with Write:
- Very specific: line 35
Read the mode in the order ${CLAUDE_PLUGIN_ROOT}/modes/README.md gives (session selection → .claude/mode → B) and copy the role set from …
- Very specific: line 49
Put the task id in the Agent description: "[T-3] SMS pause sentence", and call the role by its plugin name (subagent_type: …
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
board-plan — the session's plan file
Why this exists
When several sessions and agents work at once — across more than one project — the chat alone no longer shows who is doing what, what is waiting for you and where each commit has landed. Nothing else records the plan: which tasks exist, which role does each one, and whether a returned agent's work was actually accepted. This skill keeps that plan in one small file per session, so it survives a long session and a context compaction, and you can open it at any time.
Contributions are welcome: https://github.com/adyusuf/claude-code-standards
Standard: ${CLAUDE_PLUGIN_ROOT}/standards/22-live-board.md.
1. The file
.claude/board/plan-${CLAUDE_SESSION_ID}.md at the repository root — one file per
session, so parallel sessions never write the same file. .claude/board/ is git-ignored
(runtime data); add the line to .gitignore if it is missing, and say so.
Create it before the first piece starts, with Write:
# Plan — <short goal> (<dd/mm/yyyy>)
Mode: <letter> · Roles: <role set copied from the mode file, or "none">
| Id | Task | Role | Branch | Status | Commit | Note |
|---|---|---|---|---|---|---|Read the mode in the order ${CLAUDE_PLUGIN_ROOT}/modes/README.md gives (session
selection → .claude/mode → B) and copy the role set from that mode's file — never from
memory. Mode A has no agents: Roles: none, the file tracks tasks only.
2. Tasks
One row per unit that merges to dev on its own (#26). Ids are T-1, T-2, … in this
file; the next id is one more than the highest in it. Route by
${CLAUDE_PLUGIN_ROOT}/modes/role-selection.md; a role outside the active mode is not
started — propose the mode change in one line (#27). Work you do yourself has the role
me.
3. Tag every agent call
Put the task id in the Agent description: "[T-3] SMS pause sentence", and call the
role by its plugin name (subagent_type: "adyusuf:<role>"). The tag links the agent's
work to its row in the plan file.
4. Keep the status current — in the same turn as the work
| When | Status |
| Not started | todo |
| You or an agent started it | running |
| The agent returned, not yet audited | agent_done |
| You audited the output and the task is closed | done |
| Blocked (review, the user, another task) | waiting + what, in Note |
| Failed and not retried now | failed + why, in Note |
| A question only the user can answer | needs_decision + the question, in Note |
agent_done means "the agent returned", not "the task is done" (#28). Record each
task's commit as soon as it exists (several: <sha>, <sha>).
Asking the user: set needs_decision, ask the same question in the chat
(AskUserQuestion), and move the row on when the answer comes. An answer never stands
in for an approval that needs the user's explicit yes (deploy, deletion, sending
outward) — that is asked for in the chat in its own words. A promotion is not asked: it
follows the configuration (#34).
5. The chat table still ships
The plan file is an extra record; the status table in the reply stays mandatory. Take
its figures from the file, not from memory. The on-screen task list stays too, with the
same items: phases, decision tasks, related work inserted, unrelated work queued — never
dropped (${CLAUDE_PLUGIN_ROOT}/standards/00-working-method.md §1a).
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
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.

