TEST COVERAGE

2026-08-12 15:49:11 UTC · commit ae253bec · arrows vs be894ab5 (70.7%)every test, every gate — the rigor moat, made visible

Source line coverage

70.9% ▲ +0.2pp

Tests passing (local)

7,555 ▲ +111.0

Marker-backed

49%

BDD-style (G/W/T)

34%

Smoke set · make smoke (by fn)

833 ~16s · 80% loop

Coverage by feature area

Telegram commands	60.4% 1069/1769	±0
Bots & notify	62.6% 1306/2086	▲ +0.1pp
Fundamentals & filings	66.7% 2284/3426	▼ −2.6pp
Agent cycles	66.8% 5628/8421	▲ +0.9pp
Research & experiments	67.0% 6470/9651	▲ +0.8pp
Data feeds & ingest	67.5% 3053/4521	±0
Core & uncategorized	72.0% 16606/23068	▲ +0.2pp
Signal pipeline	72.0% 213/296	±0
Infra & scheduling	73.6% 2900/3938	▲ +1.2pp
Books & ledger	79.8% 1081/1354	±0
Voice & output	81.5% 1767/2168	±0
The Floor (Lockbox)	83.6% 2283/2731	±0
Learning & memory	85.4% 1145/1340	±0
Stream & OBS	86.9% 450/518	±0
Services	91.8% 178/194	±0

Tests by type

by function (7,296) · 49% marker-backed (ground truth), the rest bucketed by a content heuristic.

Unit	5,496	▲ +79.0
Integration	1,141	▲ +23.0
UI (Telegram/Discord)	455	±0
Contract	170	▲ +1.0
Security	34	±0

Source-run only — the "tests passing" count parses pytest's "passed" and has no CI linkage, so CI-only failures (flaky network tests) do not appear here. See RUNBOOK for the gate vs CI distinction.