docs(skills): propagate neutralizationGate (agent-eval 0.107.0) into eval/promotion skills

[drewstone]

Additive, doc-only propagation of the new footprint-matched placebo promotion gate (neutralizationGate + neutralizeText, shipped in @tangle-network/agent-eval 0.107.0 from @tangle-network/agent-eval/campaign) into the eval/promotion skill family. No code or behavior changes — markdown only.

The gate is a placebo arm: it proves a held-out lift came from the candidate's CONTENT, not from the extra prompt/mount FOOTPRINT the content added. A held-out lift that a content-blanked, footprint-matched twin reproduces (by default ≥ maxDecorativeFraction = 0.5) is decorative and must be held. It composes AFTER the significance gate — composeGate(heldOutGate({...}), neutralizationGate({...})) — and is wired via runImprovementLoop({ neutralize }), which populates ctx.neutralizedJudgeScores / ctx.neutralizedArtifacts.

Files updated (5)

• eval-campaign/SKILL.md — bumped the peer floor to >= 0.100.0 (current published 0.107.0) and added the layering-rule note; extended the config-reference gate row to describe composing neutralizationGate and the fact that selfImprove/this scaffold do not yet surface the neutralize option (wire at the runImprovementLoop level or re-export first).
• improve-conductor/SKILL.md — Invariant 1 now states a paired, significant lift is still not shippable until a footprint-matched placebo shows the gain is content, not footprint; routes the check through measurement-validation.
• measurement-validation/SKILL.md — replaced the Invariant-2 "Enforced by" note to compose trustVerdicts with the agent-eval statistical gates: powerPreflight, heldOutGate / heldoutSignificance, and neutralizationGate + neutralizeText.
• surface-evolution/SKILL.md — Invariant 2 now also gates on a footprint placebo (composeGate(heldOutGate({...}), neutralizationGate({...}))); added neutralizationGate and composeGate to the substrate-primitive list.
• skill-evolution/SKILL.md — the self-revision gate and the slot-machine anti-pattern now both name the footprint-matched placebo (neutralization) as a required check.

[tangletools]

✅ Auto-approved drewstone PR — 2347aaf3

This PR was opened by the trusted drewstone account.
The full PR reviewer audit still runs separately and will publish findings if it detects issues.

_{tangletools · auto-approval · reason: drewstone_author · 2026-07-08T00:32:54Z}