mesh/panel (draft): convene independent perspectives and return their convergence

[sergiobuilds]

Opens the mesh/ quadrant, reserved in the README for "converge independent views into consensus". Posted as a draft — the shape is deliberately narrow, so the review can settle on where mesh/ starts before more skills follow.

What it does

Draws two to four perspectives whose failure modes are distinct — for example, correctness, security, readability, cost, adversarial user — and returns their convergence: where they agree, where they disagree, and the single next question that would resolve the disagreement.

Not a checklist, not an average. Averaging is the failure mode this skill exists to prevent.

Design commitments

• Independent perspectives, not roles. A senior and a junior of the same lens count as one, not two.
• One load-bearing reason per perspective. Hedging lists are dropped.
• Disagreement is the product. A panel that agrees on everything did not need to be convened.
• No averaging. If two perspectives split, the answer is the question that resolves the split, not the midpoint.
• No cross-artifact carry. Every panel is drawn for the artifact in hand.

Fit with the other quadrants

Quadrant	Cardinality × Time	Skill's cut
depth	one artifact, now	shower, hate — one lens
breadth	many artifacts, now	ssotchk — one truth
coil	one project, across rounds	retro, flywheel — one loop
mesh	many minds, one round	panel — many lenses, one artifact

Verification

Registered in plugin.json and README's index; passes validate-skills.sh (18 skills).

Open questions for review

1. Is 2–4 the right cap on perspectives? The README's map places mesh at the "many minds" quadrant; too few and it collapses into hate, too many and it dilutes.
2. Should the output shape be a table (perspective × verdict × reason) or the current three-region grouping (agree / agree-with-different-reasons / disagree)?
3. Where does the invoker sit — model-invoked (like shower) or user-invoked (like hate)? Draft currently marks it model-invoked; happy to flip.

[sergiobuilds]

Draft → Ready. Rebased onto latest main (past the re0-* rename and the new skills) and re-verified.

What the branch carries now:

• skills/mesh/panel/SKILL.md — unchanged from the draft. Description 473 chars, four required sections (Goal / Workflow / Rules / Verification), no cross-skill name references, no ../ deep refs.
• .claude-plugin/plugin.json — appends ./skills/mesh/panel to the current 22-skill list; rest of the array untouched.
• README.md — replaces the mesh/ "in development" line with the same table shape used by depth/ / breadth/ / coil/.
• package.json — bumps 0.12.0 → 0.13.0 per the version-bump rule (new skill = minor).

Verified locally: bash scripts/validate-skills.sh → ✓ skill catalog valid (23 skills), node -e "JSON.parse(...)" on plugin.json OK.

One deferred item: docs/readme/README.*.md still carry the "in development" line for mesh/ in ten languages. Held out of this PR to keep it single-purpose and because the English README is the canonical source; happy to add a follow-up commit with the localized rows on request.

[LilMGenius]

@sergiobuilds ✨ this one opened a whole new quadrant, thank you! Your panel skill ships in v0.15.0 as prism, and here's the honest tour of what changed on the way in.

The rename: a cold-read caught that "panel" whispers consensus and vote, the very failure the skill exists to prevent (averaging the views). A prism instead fans one thing into a spectrum of independent lenses, so the name now carries the point: disagreement is the product, never an average. Bonus, it pairs cleanly with macrothink as its mirror (macrothink repeats one read for robustness, prism splits into distinct lenses for coverage).

What I tuned on top of your design:

• Flipped it to user-invoked, since plural lenses are an opt-in spend and their convergence must never pose as automatic proof (the same reason macrothink waits to be asked).
• Traded the fixed "2 to 4" for a distinctness rule: one lens per genuinely distinct failure mode, banded 2 to 5, no default. The artifact's real failure modes set the count, so nothing gets padded to a fourth or dropped at a real fifth.
• Opened mesh/ and wired it across every roster surface (plugin.json, the README Index in all 11 languages, the catalog SSOT).

The heart of it, distinct lenses returning the one question that resolves their disagreement, is all yours. Thank you again 💚