feat(model-profiles): add Fable 5.1 profile and wire context-diet to it

[paveg]

What

Distill Anthropic's Fable 5.1 prompting guide into the model-profile layer this repo already uses for model-specific facts, and wire the one consumer that was missing it. No always-on rule changes.

File	Change	Why
references/model-profiles/fable-5-1.md	new — Params / Remove / Rewrite / Keep tables, plus a section listing the guide's blocks the Claude Code harness already supplies	README's contract: a model is named in one file only; claude-md-layout and empirical-prompt-tuning resolve <session-model-id>.md
skills/context-diet/SKILL.md	"Name the failure first" loads the session model's profile	its description already promised "when a model upgrade makes old guardrails redundant", but the body never read the profile that lists them (grep model-profile hit only the description)
references/model-profiles/README.md	context-diet added to Consumers
references/japanese-writing/norms.md	AI-smell taxonomy gains Mannered prose（擬人化・比喩の重ね） with Japanese-native examples	the one guide item with no counterpart in the existing taxonomy; JIT-loaded, zero always-on cost

Source: https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/prompting-claude-fable-5-1 (fetched 2026-09-17). Three sections the secondary coverage skipped are included: append-only history / thinking-block binding, quoting retrieved sources, safeguard false positives.

Not in this PR

• No rule additions. harness-meta.md requires a Red step per rule; none of the guide's items has an observed failure here. Repo grep for the guide's Remove targets (anti-formatting rules, narration suppressors) found none.
• The profile and the taxonomy entry are unverified for behavior: no empirical-prompt-tuning or writing-proofread run against them yet.

🤖 Generated with Claude Code

[changeset-bot]

⚠️ No Changeset found

Latest commit: 4ae3021

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[paveg]

Test results

• Target commit: 4ae30212174ae9880b7ccbe07938f6fd10979708

Check	Command	Result
Lint + render + format + hook/skill-script tests	just test	✓ All checks passed!
Format on the committed file	just fmt-check	passes after just fmt realigned the profile's tables (the formatter skips untracked files, so the first check ran before git add)
Always-on cost	context-diet/scripts/measure.py	rules unchanged; no new skill description
Commit signature	git log -1 --show-signature	Good "git" signature (ED25519)

Not covered

• Behavior of the profile: no empirical-prompt-tuning or claude-md-layout run has loaded fable-5-1.md yet.
• The new norms.md taxonomy entry: no writing-proofread run on a deliberately mannered draft.
• Accuracy of the distillation beyond a single read of the source page on 2026-09-17.