vixero-council
Skill1 starsChanged 7 months agoMITView on GitHub
Convene a 6-lens council (correctness, adversary, maintainer, performance, integration, contrarian) with a judge that synthesizes a verdict. Use when the user says "convene council", "council review", "review from all angles", or needs multi-lens review before a high-stakes commit. Returns lens findings, surfaced disagreements, minority report, verdict with calibrated confidence.
In this repository11 other files it gives its agents.
- assumption-surfaceSkill
- description-craftSkill
- goal-criteriaSkill
- karpathy-codingSkill
- lean-responseSkill
- simplicity-auditSkill
- skill-compressSkill
- skill-graphSkill
- surgical-diffSkill
- token-auditSkill
- token-budget-checkSkill
The file
We split this skill into the rules it gives an agent
4 rules. About 0 are common, 4 are distinctive, and 0 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
0
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 13
Before: architecture choice, public API, migration, refactor > 500 lines, production deploy
- Distinctive: line 22
Read input. Classify mode.
- Distinctive: line 24
Run judge synthesis AFTER all six emit.
- Distinctive: line 86
If ≥ 3 lenses emit blockers → verdict is REJECT regardless.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
vixero-council
Six independent lenses review the input. Judge synthesizes. Disagreements preserved. Groupthink blocked by protocol.
Trigger on
- "convene council" / "council review"
- "review from all angles" / "attack from every perspective"
- Before: architecture choice, public API, migration, refactor > 500 lines, production deploy
Scope
- review mode: code / patch / design doc / proposal → six-lens critique
- advisory mode: decision question → six-lens recommendation
- For task-scale single-lens reviews use
surgical-diff,simplicity-audit,goal-criteria
Protocol (strict order)
- Read input. Classify mode. Record subject.
- For each lens in order, open
./personas/{lens}.md, apply, emit findings. No lens reads another's output before emitting. Anti-contamination is mandatory. - Run judge synthesis AFTER all six emit.
- Render canonical output block.
The six lenses
| # | lens | question | emoji | file |
| 1 | correctness | does it do what it claims under all inputs? | 🎯 | personas/correctness.md |
| 2 | adversary | how does a motivated attacker break it? | 🗡️ | personas/adversary.md |
| 3 | maintainer | what costs time in 6 months? | 🔧 | personas/maintainer.md |
| 4 | performance | what does it cost at stated scale? | 🔥 | personas/performance.md |
| 5 | integration | does it fit the contracts it touches? | 🧩 | personas/integration.md |
| 6 | contrarian | what if the premise is wrong? | 🎭 | personas/contrarian.md |
Severity tiers
| tier | meaning |
| blocker | ship breaks production, security, or data |
| major | high debt or user harm; ship only with tracked follow-up |
| minor | nits, style, future refinement |
Judge ⚖️
- Consolidate findings; dedupe cross-lens hits; credit strongest framing.
- Surface disagreements — do NOT flatten.
- Decide:
SHIP | SHIP WITH FIXES | REWORK | REJECT. - Confidence: 0–100, rounded to 10s. No 73%, no 87%.
- Top-3 concrete actions before ship.
- Minority report: strongest dissenting lens, named.
Output
subject: {one line}
mode: {review | advisory}
--- LENS FINDINGS ---
correctness 🎯: blockers: {...} major: {...} minor: {...}
adversary 🗡️: (same shape)
maintainer 🔧: (same shape)
performance 🔥: (same shape or "skipped — scale not stated")
integration 🧩: (same shape)
contrarian 🎭: position: {...} | objection: {...}
--- DISAGREEMENTS ---
- {lens A} X, {lens B} Y. Trade-off: {what}
--- VERDICT ⚖️ ---
decision: {SHIP | SHIP WITH FIXES | REWORK | REJECT}
confidence: {0-100, rounded to 10}
reasoning: {3-5 imperative lines}
top-3 actions:
1. ... 2. ... 3. ...
minority report: lens: {which} | position: {one line}Hard rules
- Each lens runs independently; no lens reads another's emission first.
- Emit "none" when nothing found — never invent findings.
- Judge cannot downgrade a blocker; only the raising lens can retract.
- If ≥ 3 lenses emit blockers → verdict is REJECT regardless.
- Confidence rounds to 10s; no false precision.
- Performance lens skips if scale absent — ask, don't invent targets.
- Contrarian never votes; it reframes. Judge may elevate it to minority report.
- Input < 30 lines or < 3 paragraphs → refuse: "too small, use a single lens".
- Advisory mode: lenses recommend; judge picks one path AND names second-best.
Composes with
assumption-surface → advisory mode (interpret before review)
goal-criteria → defines correctness lens target
surgical-diff · simplicity-audit → pre-filters before council
karpathy-coding → applies if council recommends code changes
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

