03 — Source / Adapter Reality

Score: Source adapter coverage = 62/100. 16/16 adapters exist and DB adapters are live and recently-run; but host triggers are excluded from the canonical universe, the KB/SOP adapter is a declared partial blind spot, and there's an unreconciled DB↔host drift.

Live evidence (2026-06-05, read-only)

• wf_adapter_run_log: 21 runs total, last started 2026-06-05 02:10:44+00, 5 runs on 06-05 → adapters DID refresh today.
• wf_scanner_run_log: 5 runs total, last run 2026-06-04 09:53:28+00, 0 runs on 06-05, all SUCCESS → the unified scanner orchestrator has not logged a 06-05 run (timer armed 04:10; see doc 07).
• workflow_scanner_registry: 6 rows SCHEDULED_NON_DOT (not promoted to DOT — owner-gated, deliberate).

Per-source-class reality

Source class	Adapter exists?	Live or stale?	Partial/complete?	Last refresh	Blind spots
DB fns / dot / iucmd / job_kind / wf	Yes (fn_dot_wf_*_v2)	LIVE (views compute now)	Complete for DB universe (373 defs)	continuous	none for DB
Host crontab	Yes (wf_host_crontab_snapshot)	Snapshot, refreshed 06-05 02:10	Partial: true schedule semantics DB-invisible	06-05	cron schedule expr not parsed → drift vs dot.cron (42 vs 7)
systemd timers	Yes (wf_systemd_timer_snapshot)	Snapshot	Partial	06-05	OnCalendar semantics not normalized
FS dot/bin	Yes (wf_fs_dot_bin_snapshot)	Snapshot	287 bin → 186 mapped	06-05	101 unmapped bin objects
FS scripts	Yes (wf_fs_script_snapshot)	Snapshot	42	06-05	classification partial
Docker containers	Yes (wf_docker_container_snapshot)	Snapshot	11	06-05	runtime→process linkage weak
KB / SOP docs	Yes (wf_kb_sop_doc_snapshot)	PARTIAL adapter	Only 2 docs enumerated	06-05	declared UNKNOWN/FP-risk blind spot
Event registry	Yes (event_type_registry 52)	LIVE	Complete	continuous	0/7 process.* active
Queue / job	Yes (job_queue via census)	LIVE	job:cut verified, 8 kinds	continuous	only 1 verified process
Process runtime obs	Yes (process_run_observation)	LIVE	REAL_RUN 0 / DRY 2 / SIM 4	continuous	no real-run observations exist
Scanner logs / timers	Yes (wf_scanner_run_log)	STALE (06-04)	n/a	06-04 09:53	no 06-05 orchestrator run

Critical reality flags

1. Adapter snapshots are LIVE-ish (06-05 02:10) but the orchestrator digest is STALE (06-04). This split means: the raw snapshots refreshed today, but the unified census run that ties them together and writes wf_*_digest_v2 last completed 06-04. The wf_*_digest_v2 TABLES are stale snapshots — read the VIEWS, never the digest tables (confirmed prior gotcha; still true). Anything quoting digest tables is reporting 06-04.
2. Host triggers excluded from the 525 trigger universe. v_trigger_universe_census = 525 is DB/dot-only. Host triggers (~77) live in v_trigger_host_unmanaged_census. A "total triggers = 525" claim is partial-shown-as-total unless paired with the host census.
3. KB/SOP adapter is the standing blind spot — only 2 docs; count is FP-risk. Any "SOP coverage" claim is unreliable.
4. Classification drift unresolved: 42 cron-DOTs declared vs 7 mapped (gap 35); 287 bin vs 186 mapped (gap 101). The drift detector exists (fn_dot_wf_classification_drift_v2) but the reconciliation is manual/owner-gated.

Adapter coverage verdict

Adapters are real and recently-run (not plan-only) — strong. But "16/16 HEALTHY" overstates completeness: 2 classes (host-trigger, KB/SOP) are partial-by-design, and the orchestrator that unifies them hasn't run today. Coverage is broad but shallow at the edges, and the freshness story is split (snapshots fresh, digests stale).