fix(ai-improve): tolerate prose-wrapped JSON from the model

[ssevera1]

Reproduced 3/3 in a row on Data-Science-Toolkit while verifying the pipeline: the selection call succeeded (context built, API responded), but Haiku sometimes reasons in prose ("Looking at this codebase, I need to identify...") instead of emitting bare JSON, despite SELECT_PROMPT's existing "ONLY valid JSON" instruction. ask_json() only tried json.loads() on the whole trimmed response, so any surrounding prose made it discard an otherwise-usable answer and skip the run entirely — safely (no bad file written), but silently.

What changed

• Stronger prompt instructions in both SELECT_PROMPT and WRITE_PROMPT: explicitly forbid reasoning/commentary, spell out that a caller parses the response programmatically
• Fallback extraction in ask_json(): if the full response isn't valid JSON on its own, scan for the first balanced {...} object anywhere in it and parse that instead of giving up
• Print more of the raw response on a genuine parse failure (800 → 2000 chars) for easier debugging next time

Same file, byte-identical across all 11 repos before this change — opening the same fix everywhere rather than just on the repo that surfaced it.

[claude]

Checked the two things that could bite here. First, the new prompt text uses correctly doubled {{/}} so both SELECT_PROMPT and WRITE_PROMPT still render through .format(). Second, the balanced-brace scan in ask_json() is not JSON-string-aware, but that only costs a missed recovery, never a bad parse: braces inside file_content that close early leave an unterminated string, so json.loads raises and the function returns None — exactly the pre-existing skip-the-run behaviour — and all downstream guards (safe_target, required-key check, SHRINK_FLOOR, the max_tokens stop-reason check) remain in front of any write. Change is one file, matches the stated intent, and bundles nothing unrelated.