Work Humanizer

Before you start: read this whole install prompt, tell me in a few lines what this skill does and what installing it involves, and wait for my go-ahead before you do anything else. Prerequisites: none for the install itself: no account, no API key, no paid service, and nothing else to install first - installing is only copying one folder, verified against the fingerprints below. Using the skill afterwards needs an agent that loads SKILL.md (Claude Code, Codex, Cursor or another agent, according to the README). The optional final mechanical sweep can be done by eye or search, or by running the bundled check.py (python check.py <file>, or standard input; Python, standard library only, local processing of a text I provide). Do not install or run any of that during this install: it is my decision, later, when I actually use the skill. Install "work-humanizer" for me. It rewrites AI-sounding work writing so it reads like a careful person wrote it, without changing what it says or commits to: sales and marketing copy, support replies, HR and legal letters, emails, docs, release notes and posts. It has two modes: rewrite, where it rewrites my draft fresh from its fact list, and interview, where it asks up to five sharp questions and builds the piece from my own notes and words, then audits every fact sentence by sentence. Repository: https://github.com/phuryn/work-humanizer (the skill is the skills/work-humanizer/ folder of the repository at the pinned commit) Pinned version: install the code at commit 0a9065791d021f3dda695f3887cb2ae4c0eacbcf exactly. Do not substitute a newer version. If this commit is unreachable, stop and tell me. 1. Clone https://github.com/phuryn/work-humanizer into a temporary folder, check out commit 0a9065791d021f3dda695f3887cb2ae4c0eacbcf, and copy the skills/work-humanizer/ folder from that checkout. Then compare every file with the fingerprints recorded for this listing on 2026-10-19: [FINGERPRINTS] - SKILL.md: SHA-256 9ad38da86922e17d523e8a02aed352bcea617bbf5333e3d800df22f178a9559d - check.py: SHA-256 d8d19e486c5fe69c1c3541ba427eb7067eeae822cff2c2b81395b6364f890d1c - tells.md: SHA-256 073c69670ac8710da552d39d190a5c026ec329e062b30c45e34546a8c6d8db09 If a file does not match, it changed since the listing was prepared: do not install it yet, tell me what you found and let me decide. 2. Read SKILL.md in full and tell me in plain words: what the skill does when used, which services it connects to, whether it reads or sends any personal information or keys, and anything that seems unusual. Stop on any red flag and tell me before going further. Confirm these points from the files: - The installed package is only the skills/work-humanizer/ folder: 3 files, 29,817 bytes in total - SKILL.md (21,927 bytes), check.py (4,437 bytes) and tells.md (3,453 bytes), the last one a list of words and expressions to watch for. The rest of the repository (evidence/, examples/, tests/, including the archived copies under evidence/skill-versions/) is not the skill and is not installed. - Nothing is sent anywhere during installation itself; the only download at install time is the repository from GitHub at the pinned commit. No account, API key or telemetry is involved in installing. - In the skill folder, no host is named and there is no network code: check.py (Python, standard library only) reads a local file or standard input and counts patterns in the text (dashes, semicolons, AI vocabulary from tells.md, transitions, sentence length); it flags, it never rewrites, and it makes no network calls (checked 2026-10-19: no urllib, requests, socket, curl or wget). - At use time the skill connects to nothing: no host, no service, no telemetry, nothing uploaded, in the skill folder. The hosts that appear elsewhere in the repository (for example openrouter.ai, api.perplexity.ai, api.openai.com) belong to the author's evidence corpus and reproduction scripts under evidence/ and to the documentation, not to the installed skill. What I give the assistant to rewrite is processed by my AI provider under my plan's terms; the skill itself sends nothing. - Disclosure, checked 2026-10-23: this listing was first stopped by an earlier red-line scan of the whole repository, then cleared by Max on 2026-10-23 after a re-scan with redline-1.2.1 returned 0 hits on the repository and on the skill folder. The earlier hit came from two example lines in the author's test corpus under evidence/inputs/, outside the installed folder: the word "cookies" there means biscuits in a fictional bakery kit text, and the supposed send command was the end of the word "sync" in a fictional release note. Neither line is code or a skill instruction. - Behaviour built into the skill: it keeps every fact and invents nothing - rewrite mode writes the piece fresh from a list of the draft's facts, at the same level of certainty; interview mode asks one round of at most five questions when the draft is missing what only the writer knows, then builds the piece from the writer's own words; and it explicitly says never to promise "undetectable". 3. When I say "yes, install", install it as a skill: copy the skills/work-humanizer/ folder (everything in it, exactly as fingerprinted) into the agent's skills directory, in a folder named "work-humanizer". 4. Verify with no network calls: frontmatter valid, files in place. 5. Report what was installed, where, and what I still need to do myself (for example: give the skill a draft to rewrite, or ask for interview mode; and, only if I want the mechanical sweep run as a script, have Python ready - running check.py is optional and is a separate step that only starts when I ask for it). GitHub is optional: if I have a GitHub account or the gh CLI, you may use it; otherwise public access is fine. Never require it unless it's in the prerequisites above. Rules: don't touch anything outside the temp folder and the install target. If anything looks off, stop and ask me.