morning-briefinglisted

# Morning Briefing — Skill Instructions ## What this skill does Produce one morning briefing that merges **six sources** into a single structured view: hal tasks (every workspace you belong to, sprint-aware), Obsidian jobsearch state, the Google Calendars your workspaces declare, and two Gmail inboxes (perso + pro) — then cross-references mails against the vault and CRM to update in-flight process status. It also runs a scoring pipeline on LinkedIn job alerts found in the perso inbox, surfaces the best 2-3 offers with fit rationale, and generates an **ordered plan du jour** as the final block. The rendered brief is read-only except for one daily-log write per hal workspace at the end of the run (Step 4), plus description-only updates to a dedicated hal task when routing idea capture out of the daily log (Step 5). Any session that reviews or cleans up this skill's daily log or hal tasks — reordering, cancelling, merging duplicates, updating descriptions — is **log-only**: it must never execute a task inline (e.g. draft a LinkedIn post, write a CR). See Step 5. Any backend that is unreachable renders a loud `⚠️ <source> DOWN — <reason>` line instead of silently omitting data. Silent omission is a critical failure. --- ## Invocation modes — interactive vs `--headless` This skill runs in two modes. **Interactive** (default, no flag) is the full behaviour described in the steps below, run inside a live session with a human present. **`--headless`** is an explicit, unattende