fix: override strategist post_checks in design_workflow() with dual-format sentinels

[crqu]

Closes #1323

Changes

• Override the strategist node's post_checks in design_workflow() to use must_contain=["### Phase 1", "### Hypotheses"], accepting either phased plans (new projects) or improvement hypotheses (existing projects) via grep -qE OR semantics
• Placement after the reads loop (L643-647) preserves the study-combined.md reads update via model_copy
• Rename and update test test_design_workflow_inherits_post_checks → test_design_workflow_overrides_post_checks to reflect the override behavior and assert ### Hypotheses instead of ### Architecture

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 84.87%. Comparing base (c2161fa) to head (e01ab7e).
⚠️ Report is 13 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1324      +/-   ##
==========================================
+ Coverage   84.83%   84.87%   +0.03%     
==========================================
  Files         231      231              
  Lines       25766    25785      +19     
  Branches     4117     4120       +3     
==========================================
+ Hits        21859    21885      +26     
+ Misses       2928     2923       -5     
+ Partials      979      977       -2     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[gx-ai-architect]

@ceo-review

[github-actions]

✅ Factory Review: KEEP

Verdict: KEEP
Reason: QA: CLEAN — 5746 tests pass, 0 failures, lint clean. Code review 7/7 PASS. Adversarial QA 6/6 criteria VERIFIED. Precheck HALT on score_direction (no eval baseline, not a code regression).

QA Analysis

Adversarial QA Report

PR: #1324
Title: fix: override strategist post_checks in design_workflow() with dual-format sentinels
Detected project type: Library (Python CLI + workflow engine)
Date: 2026-08-19

Smoke Test

Command:

uv run pytest tests/test_models.py tests/test_guards.py tests/test_runners.py -x -q --tb=short -k 'not (BobAuth or preflight_error_unchanged)'

Output: 202 passed, 3 deselected in 6.31s
Result: PASS

Test Plan

The PR overrides the strategist post_checks in design_workflow() so it uses ["### Phase 1", "### Hypotheses"] instead of inheriting the build workflow's ["### Phase 1", "### Architecture"]. This is a dual-format sentinel fix — design mode produces plans with ### Hypotheses headers, not ### Architecture.

Acceptance criteria derived from the commit:

1. design_workflow() strategist has must_contain=["### Phase 1", "### Hypotheses"]
2. build_workflow() strategist retains must_contain=["### Phase 1", "### Architecture"]
3. Build workflow is not mutated as a side effect of calling design_workflow()
4. just_plan=True variant also gets the override
5. SKILL.md export includes the correct sentinels
6. No regressions in workflow or verification tests

Feature Tests

Criterion 1: design_workflow strategist has correct dual-format sentinels

Status: VERIFIED
Command:

uv run python3 -c "
from factory.workflow.definitions import design_workflow
dwf = design_workflow()
d_strat = dwf.nodes['strategist']
print('DESIGN strategist post_checks must_contain:', d_strat.post_checks[0].must_contain)
assert '### Hypotheses' in d_strat.post_checks[0].must_contain
assert '### Phase 1' in d_strat.post_checks[0].must_contain
assert '### Architecture' not in d_strat.post_checks[0].must_contain
print('PASS')
"

Output:

DESIGN strategist post_checks must_contain: ['### Phase 1', '### Hypotheses']
PASS

Criterion 2: build_workflow strategist retains original sentinels

Status: VERIFIED
Command:

uv run python3 -c "
from factory.workflow.definitions import build_workflow
bwf = build_workflow()
b_strat = bwf.nodes['strategist']
print('BUILD strategist post_checks must_contain:', b_strat.post_checks[0].must_contain)
assert '### Architecture' in b_strat.post_checks[0].must_contain
assert '### Phase 1' in b_strat.post_checks[0].must_contain
print('PASS')
"

Output:

BUILD strategist post_checks must_contain: ['### Phase 1', '### Architecture']
PASS

Criterion 3: build_workflow is not mutated by design_workflow()

Status: VERIFIED
Command:

uv run python3 -c "
from factory.workflow.definitions import design_workflow, build_workflow
bwf1 = build_workflow()
_ = design_workflow()
bwf2 = build_workflow()
assert bwf1.nodes['strategist'].post_checks[0].must_contain == bwf2.nodes['strategist'].post_checks[0].must_contain
assert '### Architecture' in bwf2.nodes['strategist'].post_checks[0].must_contain
print('PASS: build_workflow is not mutated by design_workflow()')
"

Output:

PASS: build_workflow is not mutated by design_workflow()

Criterion 4: just_plan=True variant also gets the override

Status: VERIFIED
Command:

uv run python3 -c "
from factory.workflow.definitions import design_workflow
plan_wf = design_workflow(just_plan=True)
strat = plan_wf.nodes['strategist']
print('PLAN mode must_contain:', strat.post_checks[0].must_contain)
assert '### Hypotheses' in strat.post_checks[0].must_contain
assert '### Phase 1' in strat.post_checks[0].must_contain
print('PASS')
"

Output:

PLAN mode must_contain: ['### Phase 1', '### Hypotheses']
PASS

Criterion 5: SKILL.md export includes correct sentinels

Status: VERIFIED
Command:

uv run python3 -c "
from factory.workflow.definitions import design_workflow
from factory.workflow.skill_export import workflow_to_skill_md
wf = design_workflow()
skill = workflow_to_skill_md(wf)
assert 'Hypotheses' in skill
assert 'Phase 1' in skill
print('PASS: SKILL.md export includes dual-format sentinels')
"

Output:

PASS: SKILL.md export includes dual-format sentinels

Criterion 6: No regressions in verification or workflow tests

Status: VERIFIED
Command:

uv run pytest tests/test_verification.py -v --tb=short

Output: 40 passed in 0.17s

Command:

uv run pytest tests/ -v --tb=short -k "workflow"

Output: 821 passed, 9 skipped, 4682 deselected in 13.93s

Edge Case Tests

min_size and must_exist preserved in override

Status: VERIFIED
Command:

uv run python3 -c "
from factory.workflow.definitions import design_workflow
wf = design_workflow()
strat = wf.nodes['strategist']
assert strat.post_checks[0].min_size == 200
assert strat.post_checks[0].must_exist is True
assert strat.post_checks[0].path == '.factory/strategy/current.md'
print('PASS: all fields preserved')
"

Output:

PASS: all fields preserved

Acceptance Criteria Verification

#	Criterion	Status
1	design_workflow strategist uses ["### Phase 1", "### Hypotheses"]	VERIFIED
2	build_workflow strategist retains ["### Phase 1", "### Architecture"]	VERIFIED
3	build_workflow not mutated by design_workflow() call	VERIFIED
4	just_plan=True variant inherits override	VERIFIED
5	SKILL.md export includes correct sentinels	VERIFIED
6	No regressions in test suite	VERIFIED

Adversarial Verdict: PASS

All acceptance criteria verified with evidence. The model_copy approach correctly overrides the inherited post_checks without mutating the build workflow. Both just_plan=True and default design mode get the correct dual-format sentinels. No regressions detected across 821 workflow tests.

---

Posted by Factory CEO