satoshikawatosatoshikawato/gbdraw/.claude/skills/gbdraw-campaign/SKILL.md

gbdraw-campaign

Skill162 starsChanged 15 days agoMITView on GitHub

Keep the state of a multi-session gbdraw task in files - start, resume, hand off, report progress, and finish a campaign with STATE.md, LOG.md, DECISIONS.md, and progress.json in the task folder. Use when starting or resuming work that spans sessions or PRs, before compacting or ending such a session, and when asked how far the work has progressed.

We split this skill into the rules it gives an agent

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

≈0
Common · estimate
Many files like this have a rule of this kind.
≈0
Distinctive · estimate
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.

Common
Distinctive
Very specific

Very specific:
If the folder is new, copy templates/ into it and fill STATE.md and progress.json.
line 24
Very specific:
Read STATE.md and DECISIONS.md.
line 29
Very specific:
After each merge, abandoned approach, or decision, update STATE.md and append to LOG.md or DECISIONS.md in the same turn.
line 33
Very specific:
Before ending, update STATE.md and write NEXT-SESSION-PROMPT.md from the template: under 10 KB, pointing to the files above instead of …
line 51
Very specific:
Write the final report from DECISIONS.md, progress.json, and LOG.md.
line 57

≈ “Common” is an estimate: at least 1 in 5 of the 308,232 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.

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

Files

SKILL.md

Campaign state files

A campaign is work that spans sessions, PRs, or agents. Its files live in one task folder outside the repository ($TASK_DIR; the local instructions say where). The files carry the state, not the conversation. This skill does not decide the campaign's scope.

File	Content	Rule
`STATE.md`	Current state only: goal and done criteria, open PRs (number, branch, head SHA, CI), worktrees, the next steps, open questions, assigned ID blocks	Overwrite; at most 150 lines. Read it first after a resume or compaction.
`LOG.md`	What happened, in time order	Append only. Search it (`grep`, `tail`); never read it whole.
`DECISIONS.md`	Each Owner decision and each Owner-delegated recommendation: id, date, question, choice, source	Append only. Cite the ids in PR bodies and the final report.
`progress.json`	Work items with weight and status	Update when an item changes.

Templates are in templates/.

Start or resume

If the folder is new, copy templates/ into it and fill STATE.md and progress.json.
Bind this session to the folder, so that the SessionStart hook reloads STATE.md after a compaction or a resume: python3 .claude/skills/gbdraw-campaign/scripts/bind_session.py "$TASK_DIR".
Read STATE.md and DECISIONS.md. Search LOG.md only for what you need.

While working

After each merge, abandoned approach, or decision, update STATE.md and append to LOG.md or DECISIONS.md in the same turn.
Record each newly found bug at once under the next ID of the campaign's block, with steps, expected result, actual result, and cause.
Answer "how far along is it?" by running python3 .claude/skills/gbdraw-campaign/scripts/progress.py "$TASK_DIR" (--gh also reads PR states from GitHub), not from memory.

Split the session

End the session and continue in a new one when any of these holds:

a PR merged and the next item touches other files;
the context is past about 300k tokens;
a second compaction would be needed;
the session's goal is done;
the session sat idle for more than an hour, so the prompt cache is gone.

Before ending, update STATE.md and write NEXT-SESSION-PROMPT.md from the template: under 10 KB, pointing to the files above instead of copying them. Give the Owner its path.

Finish

Write the final report from DECISIONS.md, progress.json, and LOG.md. Mark the campaign complete in STATE.md, and remove the session bindings with bind_session.py --clear "$TASK_DIR".

Download now (ZIP)

In this repository11 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- avoid-ai-writingSkill
- gbdraw-promotionSkill
- gbdraw-ship-prSkill
- gbdraw-videoSkill
- gbdraw-worktreeSkill
- love-me-love-my-docsSkill
- translate-web-uiSkill
- web-gallery-screenshot-maintenanceSkill
- write-clear-pull-requestSkill

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

What did you use it for? e.g. Reviewing pull requests in a Rails app

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.

Check your setup, free