RomanNum3ral
/
gstack
mirror of https://github.com/garrytan/gstack.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Merge 1a632895ce into a3259400a3

This commit is contained in:
genisis0x 2026-07-14 19:16:52 -07:00 committed by GitHub
parent a3259400a3 1a632895ce
commit 83fe0651b2
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 95 additions and 12 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

27
autoplan/SKILL.md
View File

@ -1227,8 +1227,10 @@ CEO DUAL VOICES — CONSENSUS TABLE:
5. Competitive/market risks covered? — — —
6. 6-month trajectory sound? — — —
═══════════════════════════════════════════════════════════════
CONFIRMED = both agree. DISAGREE = models differ (→ taste decision).
Missing voice = N/A (not CONFIRMED). Single critical finding from one voice = flagged regardless.
CONFIRMED = both voices agree. CONFIRMED-1V = only one voice ran (other N/A); the
orchestrator concurs with the sole reviewer, so the row is NOT dual-confirmed.
DISAGREE = models differ (→ taste decision). A missing voice is N/A, never CONFIRMED.
Single critical finding from one voice = flagged regardless.
```
Sections 1-10 — for EACH section, run the evaluation criteria from the loaded skill file:
@ -1417,8 +1419,10 @@ ENG DUAL VOICES — CONSENSUS TABLE:
5. Error paths handled? — — —
6. Deployment risk manageable? — — —
═══════════════════════════════════════════════════════════════
CONFIRMED = both agree. DISAGREE = models differ (→ taste decision).
Missing voice = N/A (not CONFIRMED). Single critical finding from one voice = flagged regardless.
CONFIRMED = both voices agree. CONFIRMED-1V = only one voice ran (other N/A); the
orchestrator concurs with the sole reviewer, so the row is NOT dual-confirmed.
DISAGREE = models differ (→ taste decision). A missing voice is N/A, never CONFIRMED.
Single critical finding from one voice = flagged regardless.
```
3. Section 1 (Architecture): Produce ASCII dependency graph showing new components
@ -1540,8 +1544,10 @@ DX DUAL VOICES — CONSENSUS TABLE:
5. Upgrade path safe? — — —
6. Dev environment friction-free? — — —
═══════════════════════════════════════════════════════════════
CONFIRMED = both agree. DISAGREE = models differ (→ taste decision).
Missing voice = N/A (not CONFIRMED). Single critical finding from one voice = flagged regardless.
CONFIRMED = both voices agree. CONFIRMED-1V = only one voice ran (other N/A); the
orchestrator concurs with the sole reviewer, so the row is NOT dual-confirmed.
DISAGREE = models differ (→ taste decision). A missing voice is N/A, never CONFIRMED.
Single critical finding from one voice = flagged regardless.
```
3. Passes 1-8: Run each from loaded skill. Rate 0-10. Auto-decide each issue.
@ -1721,6 +1727,14 @@ Present as a message, then use AskUserQuestion:
```
## /autoplan Review Complete
### Review Mode
[If any review phase ran single-voice (Codex unavailable/degraded — see the
degradation matrix), show this banner; otherwise omit the whole section:]
⚠ SINGLE-VOICE MODE — Codex unavailable for [N] of [M] review phases. The
consensus columns for those phases reflect one independent reviewer, not two;
their confirmed rows read CONFIRMED-1V (orchestrator concurs with the sole
reviewer), not CONFIRMED.
### Plan Summary
[1-3 sentence summary]
@ -1771,6 +1785,7 @@ I recommend [X] — [principle]. But [Y] is also viable:
```
**Cognitive load management:**
- All review phases ran dual voices: skip the "Review Mode" banner. Any phase single-voice (Codex unavailable): show it, with N = phases that ran without Codex and M = total review phases.
- 0 user challenges: skip "User Challenges" section
- 0 taste decisions: skip "Your Choices" section
- 1-7 taste decisions: flat list

27
autoplan/SKILL.md.tmpl
View File

@ -363,8 +363,10 @@ CEO DUAL VOICES — CONSENSUS TABLE:
5. Competitive/market risks covered? — — —
6. 6-month trajectory sound? — — —
═══════════════════════════════════════════════════════════════
CONFIRMED = both agree. DISAGREE = models differ (→ taste decision).
Missing voice = N/A (not CONFIRMED). Single critical finding from one voice = flagged regardless.
CONFIRMED = both voices agree. CONFIRMED-1V = only one voice ran (other N/A); the
orchestrator concurs with the sole reviewer, so the row is NOT dual-confirmed.
DISAGREE = models differ (→ taste decision). A missing voice is N/A, never CONFIRMED.
Single critical finding from one voice = flagged regardless.
```
Sections 1-10 — for EACH section, run the evaluation criteria from the loaded skill file:
@ -553,8 +555,10 @@ ENG DUAL VOICES — CONSENSUS TABLE:
5. Error paths handled? — — —
6. Deployment risk manageable? — — —
═══════════════════════════════════════════════════════════════
CONFIRMED = both agree. DISAGREE = models differ (→ taste decision).
Missing voice = N/A (not CONFIRMED). Single critical finding from one voice = flagged regardless.
CONFIRMED = both voices agree. CONFIRMED-1V = only one voice ran (other N/A); the
orchestrator concurs with the sole reviewer, so the row is NOT dual-confirmed.
DISAGREE = models differ (→ taste decision). A missing voice is N/A, never CONFIRMED.
Single critical finding from one voice = flagged regardless.
```
3. Section 1 (Architecture): Produce ASCII dependency graph showing new components
@ -676,8 +680,10 @@ DX DUAL VOICES — CONSENSUS TABLE:
5. Upgrade path safe? — — —
6. Dev environment friction-free? — — —
═══════════════════════════════════════════════════════════════
CONFIRMED = both agree. DISAGREE = models differ (→ taste decision).
Missing voice = N/A (not CONFIRMED). Single critical finding from one voice = flagged regardless.
CONFIRMED = both voices agree. CONFIRMED-1V = only one voice ran (other N/A); the
orchestrator concurs with the sole reviewer, so the row is NOT dual-confirmed.
DISAGREE = models differ (→ taste decision). A missing voice is N/A, never CONFIRMED.
Single critical finding from one voice = flagged regardless.
```
3. Passes 1-8: Run each from loaded skill. Rate 0-10. Auto-decide each issue.
@ -784,6 +790,14 @@ Present as a message, then use AskUserQuestion:
```
## /autoplan Review Complete
### Review Mode
[If any review phase ran single-voice (Codex unavailable/degraded — see the
degradation matrix), show this banner; otherwise omit the whole section:]
⚠ SINGLE-VOICE MODE — Codex unavailable for [N] of [M] review phases. The
consensus columns for those phases reflect one independent reviewer, not two;
their confirmed rows read CONFIRMED-1V (orchestrator concurs with the sole
reviewer), not CONFIRMED.
### Plan Summary
[1-3 sentence summary]
@ -834,6 +848,7 @@ I recommend [X] — [principle]. But [Y] is also viable:
```
**Cognitive load management:**
- All review phases ran dual voices: skip the "Review Mode" banner. Any phase single-voice (Codex unavailable): show it, with N = phases that ran without Codex and M = total review phases.
- 0 user challenges: skip "User Challenges" section
- 0 taste decisions: skip "Your Choices" section
- 1-7 taste decisions: flat list

53
test/autoplan-single-voice-label.test.ts Normal file
View File

@ -0,0 +1,53 @@
/**
* /autoplan single-voice consensus labeling (gate tier)
*
* When Codex is unavailable, every review phase degrades to a single Claude
* voice. Before #1956 the Final Approval Gate still printed CONFIRMED for those
* rows, which reads as "two independent reviewers agreed" when only one ran —
* a silent quality regression. These static checks pin the generated skill so
* the degradation stays visible: a CONFIRMED-1V label plus a top-of-gate banner.
*/
import { describe, test, expect } from 'bun:test';
import * as fs from 'fs';
import * as path from 'path';
const ROOT = path.resolve(import.meta.dir, '..');
const AUTOPLAN = fs.readFileSync(path.join(ROOT, 'autoplan', 'SKILL.md'), 'utf-8');
const AUTOPLAN_TMPL = fs.readFileSync(path.join(ROOT, 'autoplan', 'SKILL.md.tmpl'), 'utf-8');
describe('autoplan single-voice consensus labeling', () => {
test('the consensus legend defines a distinct single-voice label', () => {
expect(AUTOPLAN).toContain('CONFIRMED-1V');
// It must be explicit that one voice is NOT a dual confirmation.
expect(AUTOPLAN).toMatch(/NOT dual-confirmed/);
});
test('a missing voice is never silently rolled into CONFIRMED', () => {
expect(AUTOPLAN).toMatch(/missing voice is N\/A, never CONFIRMED/);
});
test('the Final Approval Gate carries a single-voice banner', () => {
expect(AUTOPLAN).toContain('SINGLE-VOICE MODE');
// The banner must explain WHY the consensus columns are weaker.
expect(AUTOPLAN).toMatch(/one independent reviewer, not two/);
});
test('the banner is gated on degradation, not always shown', () => {
// Cognitive-load rule: skip when every phase ran dual voices.
expect(AUTOPLAN).toMatch(/All review phases ran dual voices: skip the "Review Mode" banner/);
});
test('all three phase legends (CEO, eng, design) carry the label', () => {
// The legend block repeats once per dual-voice phase; every copy must
// teach the same CONFIRMED-1V semantics so no phase regresses alone.
const occurrences = AUTOPLAN.split('CONFIRMED-1V = only one voice ran').length - 1;
expect(occurrences).toBeGreaterThanOrEqual(3);
});
test('template and generated skill stay in sync on the labeling', () => {
for (const needle of ['CONFIRMED-1V', 'SINGLE-VOICE MODE', 'one independent reviewer, not two']) {
expect(AUTOPLAN_TMPL).toContain(needle);
}
});
});