li-humanlisted
Install: claude install-skill reportedguy/li-stack
# li-human
The humanizer. Every other writing skill runs its draft through this before
the user sees it. Rules it enforces: `../../shared/rules.md` §3-4.
**What it's for:** removing what human readers and LinkedIn's "AI slop" report
button react to (stock vocabulary in clusters, reveal bridges, staccato
rhythm, announced candor, stacked contrasts).
**What it's not for:** beating GPTZero, Pangram or Turnitin. No edit reliably
does that, and on a 200-word post their scores are noise. Never tell the user
a draft is "undetectable".
## Modes
| Mode | Does |
|---|---|
| default | clean + flag + rewrite flagged lines + score |
| `--audit` | flag + score only, no rewrite. Pre-publish check on a finished draft |
| `--profile` | read 3-6 of the user's own posts and fill or update `~/.claude/linkedin/voice.md` §5-6 |
| `--kind comment`, `reply`, `dm` | any mode, for short text: sets the length target and skips post-only checks. Passed straight to both scripts |
## The two scripts (in this skill's folder, no dependencies)
Run them by full path from this skill's base directory (the folder this
SKILL.md is in). On Windows the command is `python` (`python3` is often the
Microsoft Store stub); `py` also works. The scripts print UTF-8 themselves,
so ₹ and emoji are safe in any console.
```bash
python humanize.py draft.txt -o clean.txt --report # mechanical fixes + flags
python humanize.py draft.txt --check # flags only, change nothing
python detect.py draft.txt clean