feat(v1.10.0): "Name to fit in" naming convention (detect-and-conform)

[Filip-Podstavec]

Summary

Adds an explicit naming convention to the stack — the gap Filip flagged: the repo guided code shape (Write less, fit in) but said almost nothing about how identifiers should be named, beyond one implicit clause. Naming is one of the highest-signal long-term-maintainability factors, and AI agents have two repeatable failure modes here.

What it closes

• Casing drift — a model carries a language-default prior (Python → snake_case, JS → camelCase) and silently imposes it on a client repo that deviates, so new code reads as "the AI-written part."
• Wrong granularity — names land at the wrong altitude: too vague (get(), data()) or implementation-leaking and verbose (getting_data_from_mobile()).

Design: detect-and-conform, not a house style

Casing/separator is treated as repo-specific — the agent must detect the existing convention (per kind, per local module) and conform, never impose its default. Granularity/clarity is a universal quality principle. A prescribed house style was explicitly rejected: it would contradict the "fit in" north star and be wrong in most client repos. Full rationale + rejected alternative in ADR 0010.

Touchpoints

File	Change
AGENTS.md	tight Name to fit in bullet in "Write less, fit in" (kept under the 8 KiB lean budget)
templates/AGENTS.md.example	same bullet, self-contained (the shipped client-repo artifact)
docs/conventions.md	new ## Naming section — detection mechanics + granularity guidance
docs/adr/0010-*.md + index	the decision record
CHANGELOG.md	[1.10.0] Added entry
plugin.json / marketplace.json	version bump 1.9.0 → 1.10.0

Checks

• ✅ Version sync 1.10.0
• ✅ AGENTS.md 7.97 KiB (under the 8 KiB target — tight, as ADR 0009 predicted at saturation)
• ✅ Context-map rebuilt
• ✅ pytest tests/ — 132 passed, 0 failed