← ClaudeAtlas

daily-brieflisted

Collate every decision the chain is waiting on into one file the human reads once a day, then write their answers back. Invoked by hand, once a day. Use for "build the brief", "apply the brief", "what needs me today", "what is waiting on me".
roufmlp/agent-skills · ★ 0 · AI & Automation · score 75
Install: claude install-skill roufmlp/agent-skills
# Daily brief One file, once a day, thirty minutes. Everything the chain is waiting on arrives in it; the human's answers go back out from it. Nothing in the chain ever stops to ask the human directly — `/harden-issues`, `/run-issues` and `/parallel-hunt` default and queue rather than ask, and so does whatever feeds them issues upstream. **The brief is a view, never a store.** It is regenerated from source every run and is authoritative only between being written and being applied. Anything worth keeping lives in the issue file, the ledger or the merge briefing. A second copy of state goes stale the moment it is written. ## One command, two halves Invoked with no argument: **apply first, then rebuild.** Applying an edited brief before regenerating is what stops an answer being overwritten by the next collation. The bare invocation is the daily ritual: the human sits down, runs it, reads what comes back, edits it, and runs it again tomorrow. Yesterday's answers go out and today's brief comes in, in that order, from one command. `build` and `apply` run one half each, for when something needs redoing — a brief built against the wrong repo list, an apply that stopped halfway. Also the split to use if this is ever put on a timer: rebuilding is read-mostly and safe to automate, while applying can merge to main and deploy, so it stays with the human. State lives in `~/.claude/daily-brief/`: - **`brief.md`** — the file the human reads and edits. - **`repos.md`** — one repo pa