feat(review): add advisory simplicity lane #128

Merged

pdurlej merged 1 commit from codex/simplicity-review-lane into main

2026-06-23 19:02:38 +02:00

codex commented

2026-06-23 19:00:22 +02:00

Collaborator

Issue: #126
Value: quality
Effort: medium
Lead time: same-day

Summary

add disabled review_lanes.simplicity_review to policies/platform.v0.toml
keep default lane selection stable by choosing the single enabled lane when no --reviewer-lane is passed
add lane-aware prompt guidance for the D25 simplicity finding vocabulary
add review-quorum --advisory-lane so simplicity findings can be recorded without blocking required clean lanes

Verification

PYTHONPATH=src:. python3 -m unittest tests.test_policy_bundle tests.test_cli_render_review_prompt tests.test_cli_review_run tests.test_review_quorum tests.test_cli_review_quorum
PYTHONPATH=src:. python3 -m unittest tests.test_artifact_schema_contract tests.test_redrive_plan tests.test_cli_redrive_plan tests.test_policy_bundle tests.test_cli_render_review_prompt tests.test_cli_review_run tests.test_review_quorum tests.test_cli_review_quorum
PYTHONPATH=src:. python3 -m unittest discover -s tests
git diff --check

Boundary: no Ponytail runtime dependency, no hooks/adapters, no schema bump, no blocker promotion.

Refs #126

Issue: #126 Value: quality Effort: medium Lead time: same-day ## Summary - add disabled `review_lanes.simplicity_review` to `policies/platform.v0.toml` - keep default lane selection stable by choosing the single enabled lane when no `--reviewer-lane` is passed - add lane-aware prompt guidance for the D25 simplicity finding vocabulary - add `review-quorum --advisory-lane` so simplicity findings can be recorded without blocking required clean lanes ## Verification - `PYTHONPATH=src:. python3 -m unittest tests.test_policy_bundle tests.test_cli_render_review_prompt tests.test_cli_review_run tests.test_review_quorum tests.test_cli_review_quorum` - `PYTHONPATH=src:. python3 -m unittest tests.test_artifact_schema_contract tests.test_redrive_plan tests.test_cli_redrive_plan tests.test_policy_bundle tests.test_cli_render_review_prompt tests.test_cli_review_run tests.test_review_quorum tests.test_cli_review_quorum` - `PYTHONPATH=src:. python3 -m unittest discover -s tests` - `git diff --check` Boundary: no Ponytail runtime dependency, no hooks/adapters, no schema bump, no blocker promotion. Refs #126

codex added the

labels

2026-06-23 19:00:22 +02:00

codex added 1 commit

2026-06-23 19:00:22 +02:00

feat(review): add advisory simplicity lane

fallow-py / fallow-py-advisory (pull_request) Successful in 16s

Details

37ffb8c086

pdurlej merged commit 101f398bc1 into main

2026-06-23 19:02:38 +02:00

pdurlej referenced this pull request from a commit

2026-06-23 19:02:39 +02:00

Merge pull request 'feat(review): add advisory simplicity lane' (#128)

claude commented

2026-06-23 19:04:57 +02:00

Collaborator

🏛️ Architect review (loop) — ✅ sound, faithful to D25

Verified the subtle correctness property that makes "advisory" actually safe:

Non-blocking: _lane_entry sets status = "clean" for an advisory lane before the elif findings: status = "blocked" branch (review_quorum.py:161). So an advisory lane can never reach status == "blocked", and the verdict aggregation (if any(lane["status"] == "blocked" …)) never trips on it. ✅
Non-inflating: advisory lanes don't pad the clean count either — clean_count is summed over quorum_lanes (required-only), and the required ∩ advisory = ∅ guard (review_quorum.py:42-45) ensures an advisory lane is never in that set. So it can't falsely satisfy min_clean. ✅
Still recorded: the lane entry + its findings persist, just non-blocking — exactly D25's "record for dogfood, don't block" intent. ✅

Also confirmed: lane is enabled = false / required_for = [] / fail_on_missing = false (default-off), and the change touches only policy/prompt/quorum — no schema bump, no forgejo_client/controller/executor surface, no new dependency. 601 tests green. D25 boundary (no Ponytail runtime dep, no blocker promotion) honored.

On cloud_review = "allowed": withdrawn as a concern — the lane is intrinsically a cloud lane (model = kimi-k2.6:cloud), so allowed is internally coherent, and it's default-off, so nothing egresses until an operator knowingly opts in (with D23's guard still clamping sensitive classes). No change requested.

Clean execution of the governance decision. — claude (architect loop)

## 🏛️ Architect review (loop) — ✅ sound, faithful to D25 Verified the subtle correctness property that makes "advisory" actually safe: - **Non-blocking**: `_lane_entry` sets `status = "clean"` for an advisory lane *before* the `elif findings: status = "blocked"` branch (review_quorum.py:161). So an advisory lane can never reach `status == "blocked"`, and the verdict aggregation (`if any(lane["status"] == "blocked" …)`) never trips on it. ✅ - **Non-inflating**: advisory lanes don't pad the clean count either — `clean_count` is summed over `quorum_lanes` (required-only), and the `required ∩ advisory = ∅` guard (review_quorum.py:42-45) ensures an advisory lane is never in that set. So it can't falsely satisfy `min_clean`. ✅ - **Still recorded**: the lane entry + its findings persist, just non-blocking — exactly D25's "record for dogfood, don't block" intent. ✅ Also confirmed: lane is `enabled = false` / `required_for = []` / `fail_on_missing = false` (default-off), and the change touches **only** policy/prompt/quorum — no schema bump, no `forgejo_client`/controller/executor surface, no new dependency. 601 tests green. D25 boundary (no Ponytail runtime dep, no blocker promotion) honored. On `cloud_review = "allowed"`: withdrawn as a concern — the lane is intrinsically a cloud lane (`model = kimi-k2.6:cloud`), so `allowed` is internally coherent, and it's default-off, so nothing egresses until an operator knowingly opts in (with D23's guard still clamping sensitive classes). No change requested. Clean execution of the governance decision. — claude (architect loop)

claude referenced this pull request

2026-06-23 23:02:12 +02:00

Add Ponytail anti-slop simplicity review module #155