Skip to content

2.1.2: lists and tables in replies when the content has that shape

Latest

Choose a tag to compare

@AminBlg AminBlg released this 10 Nov 17:18
300432a
  • Changed: the reply rule. A reply is prose by default. A label line, a list,
    or a table is legal when the content has that shape: separate blocks, three
    or more parallel items, steps in order, or a comparison across attributes.
    Bold stays banned.

  • Changed: the Stop hook no longer reports headers and list items. It still
    reports bold, em-dashes, slop words, openers, and closers.

  • Changed: evals/run_reply_bench.py defaults to claude-sonnet-5-5.

  • Tested: the same reply rule ran once through evals/run_reply_bench.py on
    claude-sonnet-5-5 at low effort, 2026-11-10, inside the 2.2.0 rule block.
    The replies used lists only for steps, options, or checks, and used no bold
    or headers. Two replies used a two-item list, which the rule does not allow.
    This release carries the 2.1.1 document rules, so that run is not a test of
    this exact prompt. Its files are not committed. No published README figure
    covers 2.1.2.

  • Known gap: reader_check in evals/ste_lint.py still counts headers and
    list items in visible_total, so the reply benchmark scores a fitting list
    as a defect.

  • Rechecked after the changelog was written: the reply benchmark ran on this
    exact prompt against the v2.1.1 prompt, claude-sonnet-5-5, low effort,
    2026-11-10, one reply per prompt. No reply used bold or headers, and every
    list held three or more items.