Caveman

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 to install this skill — no account, no key, no payment, and no script or other code in its folder. One behaviour to know before using it: this is a persistent response style. Once invoked, it stays on for every response of the whole session until you say "stop caveman" or "normal mode" — that persistence is the point of the skill, and it is disclosed here rather than discovered later. Install "Caveman" for me. It switches the agent into an ultra-concise "caveman" voice: answer first, all ceremony gone (no greeting, hedging, pleasantries, recap or closer), short words, articles dropped only when the sentence still reads in one pass, one idea per sentence — while every technical fact survives and code, commands, paths, numbers and exact error strings are kept verbatim, because a dropped negation costs more than every token saved. It is a voice, not broken grammar, and it must never perform caveman (no "me think", no decorative tables or emoji). It deliberately breaks back into plain prose for security warnings, irreversible-action confirmations, step sequences a fragment could scramble, a confused user, questions to the user, and anything persisted outside the chat. Its README is honest about the payoff: measured once on ten prompts, the base caveman voice came out about 3% fewer output tokens than a plain "Answer concisely." control, inside the noise (its stricter sibling modes measured more). The /caveman ultra and /caveman wenyan aliases belong to its sibling skills ultracave and megacave, which are separate installs and are not included here. Repository: https://github.com/juliusbrussee/caveman (the skill is the whole skills/caveman/ folder at the pinned commit: SKILL.md and README.md only, 2 files in total — the canonical skills/caveman/ folder; the duplicate copy under plugins/caveman/skills/caveman/ inside the same repository is NOT part of this install) Pinned version: install the code at commit bad48a0e1df55e05e3f52249afa35ecc1909ee76 exactly. Do not substitute a newer version. If this commit is unreachable, stop and tell me. 1. Clone https://github.com/juliusbrussee/caveman into a temporary folder, check out commit bad48a0e1df55e05e3f52249afa35ecc1909ee76, and copy the skills/caveman/ folder from that checkout. Then compare every file in the folder with the fingerprints recorded for this listing on 2026-10-22: [FINGERPRINTS] - README.md: SHA-256 e8b22f37ccb204bd00cdf98c723c0df37aeda1f5863e9ad1d634f5c839d99e5d - SKILL.md: SHA-256 415d43518f2b1a9498c15a445d4657117b129ee5e81b19454ce348439cc6672b 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, what it connects to, whether it reads or sends any personal information or credentials, and anything that seems unusual. Stop on any red flag and tell me before going further. Confirm these points from the files: - The skill folder holds 2 files only: the 4,678-byte SKILL.md and its 3,030-byte README.md, from the canonical skills/caveman/ folder. The duplicate copy of this skill under plugins/caveman/skills/caveman/ in the same repository is not part of this install and its files are not in the fingerprints above. There is no script and no other code in the folder. - Nothing is sent anywhere during installation itself; the only download at install time is the skill folder from GitHub at the pinned commit. The skill reads no personal information and no credentials, at install time or at use time: it is a response style applied inside my own conversation. - The persistence is real and disclosed in the prerequisites: once invoked at use time, the mode stays on for every response of the whole session until I say "stop caveman" or "normal mode"; unsure whether it is still on, it is. Its rules keep code, commands, paths, numbers and errors verbatim, never drop negations, answer in my language, and break into plain prose for security warnings, irreversible actions, scrambled-risk step order, my questions, and anything persisted outside the chat. - Do not activate caveman mode, do not answer in caveman style, and do not use the skill in any way during this install: using it is a separate step that only starts when I invoke it later. Do not set any host configuration the README describes (such as a .caveman.json default mode) during this install — that is an optional host-level choice for me to make separately. 3. When I say "yes, install", install it as a skill: copy the whole caveman folder into the agent's skills directory, in a folder named exactly "caveman", keeping its 2 files in their original relative layout (SKILL.md at its root). 4. Verify with no network calls: frontmatter valid (name: caveman), the 2 files in place with matching fingerprints. Do not run anything from the folder and do not use the skill to verify. 5. Report what was installed, where, and what I still need to do myself (for example: that the caveman voice, once I invoke it, persists for the whole session until I say "stop caveman" or "normal mode", and that its stricter sibling modes (ultracave, megacave) are separate skills not included in this install). 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.