Consulting Integrations Sync

Workspace: use the selected project and its AGENTS.md; keep existing entity folders and Reality headings. Business paths below are relative to that project. Bundled resources are relative to this installed skill; sibling capabilities resolve by their installed names. Never search another private checkout for missing inputs. Use workspace identity, audience, pricing, and DESIGN.md fonts. Local _work adapters and evals are optional workspace tools, not bundled dependencies. Check presence and current help first; otherwise use an available connector for the same scoped operation. If neither exists, report that step incomplete. For missing scorers, perform the stated checks and label the result manual/unscored; never invent a numeric score or successful provider action.

Keep integrations/ (Granola, Attio, LinkedIn, Gmail) and the curated repo from going stale. Run end-to-end, then report. Use the connected account or explicitly selected workspace environment; never search parent directories for keys.

Source boundary

Apply the selected workspace's intake policy to every channel before persisting content. A contact, domain, or channel match is a routing hint, not permission to capture unrelated employment or personal material embedded in the same thread. Hold mixed or ambiguous records; do not leak their titles or excerpts into skip reports. Private source capture never grants permission to publish a shared snapshot.

Steps

1. Attio (query live — system of record). Per integrations/attio/crm-sync.md, query deals via REST (POST /v2/objects/deals/records/query). For each deal, compare its stage to its filesystem location. On mismatch: move the folder (pipeline/** ↔ clients/**), fix the deal AGENTS.md status line, and regenerate pipeline/_board.md. Attio wins on stage; repo wins on artifacts.
2. Granola (review before capture). Use an available connector, or the selected workspace's list_new_notes.py, with an overlapping window (default: three days before the last successful checkpoint). Review the title and content against the workspace's client/deal scope and exclusions. Attendees alone never establish scope. Skip unrelated employment, personal material, and excluded internal meetings. Hold ambiguous or mixed records; do not save their bodies or identifying details in skip logs. For a reviewed in-scope note, a compatible local adapter is: python3 integrations/granola/_work/pull_transcript.py --note <id> --reviewed-scope --out <dest>. Destinations are clients/<client>/meetings/transcripts/<date>-<slug>.md or pipeline/<stage>/<deal>/meetings/transcripts/<date>-<slug>.md. Otherwise use a connector to fetch that one reviewed note and preserve the same destination and evidence labels. No bulk fallback. Label summaries as non-evidence and keep transcript attribution tied to the actual recording owner. Deduplicate by source ID; checkpoint completed work only. Report skipped counts without private titles, names, or excerpts. New in-scope prospects route through recoup-internal-consulting-lead-intake.
3. LinkedIn (pull signal). Refresh follower/engagement snapshots if stale; if new engagement exists, chain into recoup-internal-consulting-linkedin-audience. Update integrations/linkedin/_work/LAST_SYNCED.
4. Gmail (capture FULL threads for real relationships — Attio-gated). Scope = people who matter: query Attio live for Customers + Warm Leads + Target Accounts + open-pipeline contacts, collect their email domains/addresses (exclude any segments the workspace marks out of scope). For each client/deal, archive the full thread history (every message, untruncated): python integrations/gmail/_work/export_thread_bodies.py --query "from:<domain> OR to:<domain>" --title "<Name>" --out <clients|pipeline>/<entity>/emails/email-archive-<YYYY-MM-DD>.md. Then layer the triage on top — pull_threads.py for awaiting-my-reply — as derived signal, not a replacement for the archive. A human, non-automated sender NOT in Attio → don't archive; flag it in the report as a possible new lead. Never archive the whole mailbox (automated/no-reply/newsletters excluded). Stamp integrations/gmail/_work/LAST_SYNCED. No auth → skip + note.
5. Slack (if auth configured — log + channel, like Gmail). Pull deal-tied conversations only (never the whole workspace), and pull each one complete — every top-level message AND every thread reply (--days 0 --threads); a bounded window silently drops new replies on older threads, so don't use one. Run integrations/slack/_work/list_access.py to confirm read access, then for each row of the Deal channel routing table in integrations/slack/AGENTS.md run integrations/slack/_work/pull_conversation.py --channel <id> --token user --days 0 --threads --out <dest>/<YYYY-MM-DD>-<channel>.md. File each keeper next to its client, mine the internal product DM into knowledge/product/, and flag threads awaiting my reply. Mine only the lines new since integrations/slack/_work/LAST_SYNCED (so insights don't duplicate), then stamp it. No SLACK_* keys set → skip and note it.
6. Research wiki (read-only content source). Mined by recoup-internal-consulting-research-miner (delta-guarded via integrations/research/_work/LAST_MINED) and pulled at draft time by recoup-internal-consulting-content-drafter — see integrations/research/AGENTS.md. The Friday harvest lives in recoup-internal-consulting-friday-review; trigger the miner here too if the wiki's log.md advanced since LAST_MINED.
7. Mine fresh material. New won deals → recoup-internal-consulting-case-study-builder; recurring answers → knowledge/faqs/; reusable insights/ideas → atomic signals in signals/.
8. Stamp + report. Update each integration's LAST_SYNCED, the crm-sync "Last full reconcile" note, and write a short report: what changed · what drifted · what's stale · what needs me.

Confirm before inventing client facts. Do the routine filing/reconciling without asking.