li-human

Two tools live in this folder and they both actually run. Use them. Do not eyeball this.

python3 humanize.py draft.txt --report        # clean it, show what changed
python3 detect.py draft.txt                  # score it, five checks
python3 detect.py before.txt after.txt       # prove the delta

Both read slop.json, the lexicon: 100+ stock words and phrases with plain-English replacements, invisible character classes, typographic substitutions, and structural tells. It is meant to be edited. If the user has a word they genuinely use that the lexicon strips, remove it from the file rather than fighting the tool.

What gets fixed automatically

1. Invisible characters. Zero-width spaces and joiners, word joiners, soft hyphens, byte-order marks, Unicode tag characters, non-breaking and narrow spaces. A keyboard does not produce these. They survive copy-paste, they are invisible in every editor, and they are the single most mechanical tell in generated text. humanize.py deletes every one, including any remaining Unicode format character it does not have a name for.

2. Typography. Em dash to comma, en dash to hyphen, curly quotes to straight, ellipsis to three dots, bullet character to hyphen. The em dash pass is the one that matters: it collapses — to , and then cleans up the doubled punctuation that leaves behind.

3. The lexicon. 100+ stock words and phrases with plain-English replacements: delve, leverage, robust, seamless, crucial, "testament to", "in today's fast-paced world", "let that sink in". Capitalisation is preserved and URLs are never touched. Edit slop.json; it is yours.

What gets flagged instead of fixed

"It's not just X, it's Y" constructions, rule-of-three triads, one-word rhetorical questions, hashtag walls, reflex engagement bait, uniform sentence length. Changing the shape of a sentence needs judgement, so those are handed back with a rewrite suggestion rather than mangled by a regex.

The five checks

detect.py scores the draft 0-100 on each check; higher is more human:

The gate

Every other skill in this pack routes its output through here before the user sees it. The receipt line reports the result:

humanizer: 6 artefacts stripped, human score 84 PASS

Below 60 is a FAIL: rewrite the flagged sentences by hand (the tool shows you which) and re-run. Do not show a FAIL draft to the user as finished. Between 60 and 75 is a warning: show the draft, but name the weakest check. 75 and up ships.

When the agent cannot run scripts

Some chat interfaces cannot execute Python. In that case, apply the three automatic passes by hand using slop.json as the reference, and say so in the receipt: humanizer: manual pass (no script runtime), score n/a. Never claim a scripted score you did not compute.