Centralize all credentials in Postgres vault
→
next scheduled-task creation touches brain: confirm the standard Step 0 bootstrap block in `references/scheduled-task-bootstrap.md` loads brain.py; the vault lookup path is already inside brain.py so no extra bootstrap needed for credentials.
Auto-capture tokens into credentials vault
→
Audit all credential references in skills and references/ folder
Diagnose Postgres connectivity drops
→
Fix Postgres unreachable during scheduled task runs (recent_campaign_snapshots, ad_metrics_rolling, brain write-backs all skipped)
Publish Q2 hiring JDs
→
Post Data Mgr + Setter #1 JDs (6 days overdue, blocks Apr 17 hire date)
Trim skill descriptions for token efficiency
→
Trim 22 skill descriptions >100 words (ghl-cv-sync 155w, xlsx 152w, ghl-bot-qa 142w, meeting-transcript-to-brain 139w, ghl-api 135w, ghl-convo-audit 135w)
Align CLAUDE.md with current architecture
→
Align global vs project CLAUDE.md memory-file instruction (global says all 5, project says 3 — which 3?)
AI CV fleet-wide rollout across 28 clients
→
**update `/sessions/dreamy-epic-fermi/mnt/.claude/skills/ghl-cv-sync/SKILL.md`** — add pointer near the "17 canonical CVs" list: "ai_* CVs are out of scope — see `references/ghl-ai-prompt-templates.md` and the ghl-api skill for the ai_* PUT path."
Meta App Review submission & demos
→
**Record screen demos** per the SOW Phase 2 script once the use-case write-up is finalized `[awaiting-akash]`