- pyproject.toml: 0.6.0 -> 0.7.0
- CHANGELOG.md [0.7.0] dated 2026-04-24:
  * Added: st-fix weighted auto-selector (claims, score, false/partial,
    length-fit, subject coverage, default-AI bonus, fixed penalty);
    --include-fixed and hidden --weights flags; _generated_by stamping;
    audit table with Score/Claims/False/~Fls/Cover/Fit/Prio/Why columns
    aligned to st-verdict's lens vocabulary; recovery messages.
  * Added: cross_st/_report_signals.py shared module (parse_prompt,
    parse_claims, collect_claims, get_prompt_text, verdict_normalise,
    calendar_context, report_tokens, CLAIM_BLOCK_RE) — st-verdict,
    st-fix, and st-fact all import the same regex/normalisation rules.
  * Added: st-fact per-story progress header and self-describing tqdm
    bar; --quiet/--silent suppress; also fixes the s:None per-story
    report-prefix annotation.
  * Fixed: anthropic structured claims silently dropped (Claims=- in
    st-ls). Relaxed CLAIM_BLOCK_RE to tolerate stray *_ markup in any
    whitespace gap; consolidated st-fact to import from _report_signals.

- docs/wiki/st-fact.md:
  * "Run before" now also lists st-verdict and st-cross.
  * Options table: added --retry-budget row.
  * For developers: noted the shared CLAIM_BLOCK_RE / anthropic regex
    fix so future readers know the structured claims now match.

- cross_st/_report_signals.py: typo in comment
  (st-internal -> cross-internal).

Tests: 816 passed / 114 skipped / 0 failing.