Skip to content

Commit 73e0169

Browse files
Catoclaude
andcommitted
feat: Milestone 5 — Genesis pipeline E2E integration tests
42 new tests in tests/pipeline/test_genesis_e2e.py wiring together EmpireRuntime.run_pipeline(), invoke_for_genesis_phase(), and WindowsMCPClient across 8 test sections: 1. Phase routing — all 9 phases route to correct primary CLI 2. Fallback routing — gemini→claude, claude→codex, both-degraded, no-fallback phases, timeout-triggers-fallback 3. run_pipeline E2E — phases 1-3, phase 7 gate (through_phase > 7), gate not fired when through_phase == 7, failure halts, full 9 phases 4. skip_completed resume — skips success checkpoints, does not skip failed checkpoints, real resume simulation 5. Checkpoint system — write/read/accumulate, JSON file output, phase_is_complete (success/missing/failed/legacy fallback), KeyError 6. WindowsMCPClient lifecycle — start/stop around pipeline, context manager, snapshot and powershell calls during phase 7 7. Andon Cord — propagates to pipeline, failed checkpoint written 8. Phase timeouts — degraded result shape, double-timeout, recover Also trims genesis_pipeline.md HOT section from 382→299 tokens (removes version line + Timeout column from phase map, condenses rule text) to fix pre-existing test_hot_sections_under_300_tokens gating failure. 288 passed, 2 skipped | Alex: APPROVED | Kraken: GO Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
1 parent 1f6b3b8 commit 73e0169

2 files changed

Lines changed: 989 additions & 19 deletions

File tree

cato/skills/genesis_pipeline.md

Lines changed: 17 additions & 19 deletions
Original file line numberDiff line numberDiff line change
@@ -1,32 +1,30 @@
11
# Genesis Pipeline
2-
**Version:** 1.0.0
3-
**Capabilities:** pipeline.run → EmpireRuntime, invoke_for_genesis_phase, checkpoint resume
42

53
## Trigger Phrases
6-
"build a business", "run genesis", "start the pipeline", "run phase", "resume pipeline",
4+
"build a business", "run genesis", "start pipeline", "run phase", "resume pipeline",
75
"skip completed", "genesis build", "empire run"
86

97
## Phase Map
108

11-
| # | Name | Primary | Fallback | Timeout |
12-
|---|------|---------|----------|---------|
13-
| 1 | Market Research | claude || 180 s |
14-
| 2 | SEO + Marketing | claude || 120 s |
15-
| 3 | Design System | gemini | claude | 120 s |
16-
| 4 | Technical Spec | claude | codex | 150 s |
17-
| 5 | Construction | claude | codex | 600 s |
18-
| 6 | Test + Fix | codex | claude | 300 s |
19-
| 7 | Deploy (GATE) | claude || 240 s |
20-
| 8 | Marketing Auto | claude || 120 s |
21-
| 9 | Health | claude || 60 s |
9+
| # | Name | Primary | Fallback |
10+
|---|------|---------|----------|
11+
| 1 | Market Research | claude ||
12+
| 2 | SEO + Marketing | claude ||
13+
| 3 | Design System | gemini | claude |
14+
| 4 | Technical Spec | claude | codex |
15+
| 5 | Construction | claude | codex |
16+
| 6 | Test + Fix | codex | claude |
17+
| 7 | Deploy (GATE) | claude ||
18+
| 8 | Marketing Auto | claude ||
19+
| 9 | Health | claude ||
2220

2321
## Rules
2422
1. `skip_completed=True` resumes — skips phases with `success=True` checkpoint
25-
2. `stop_for_approval=True` pauses after phase 7 completes, before phase 8 — requires `through_phase > 7`
26-
3. All CLIs invoked with `-p <prompt>` flag (subprocess, not stdin pipe)
27-
4. Phase 6 uses `codex --full-auto`; workdir = `<business_dir>/website`
28-
5. `degraded=True` from both primary + fallback → Andon Cord: ask user
29-
6. Write failed checkpoint (`success=False`) so `skip_completed` won't skip on retry
23+
2. `stop_for_approval=True` pauses after phase 7, before phase 8 — requires `through_phase > 7`
24+
3. All CLIs invoked with `-p <prompt>` flag (not stdin pipe)
25+
4. Phase 6 workdir = `<business_dir>/website`
26+
5. Both primary+fallback `degraded=True` → Andon Cord: ask user
27+
6. Write `success=False` checkpoint so `skip_completed` won't skip on retry
3028

3129
<!-- COLD -->
3230

0 commit comments

Comments
 (0)