$npx -y skills add comol/ai_rules_1c --skill cavemanUltra-compressed communication mode. Cuts output tokens ~65–75% by using a terse "caveman" style while keeping full technical accuracy. Active by default for ALL tasks (.dev.env CAVEMAN=on, the default). CAVEMAN=auto restricts it to development tasks and turns it off for an
| 1 | # caveman — terse output style |
| 2 | |
| 3 | Adapted from https://github.com/JuliusBrussee/caveman (MIT). Compress prose. Keep substance. Brain big, mouth small. |
| 4 | |
| 5 | ## Scope — where caveman applies |
| 6 | |
| 7 | Under the default **`CAVEMAN=on`**, caveman is active for **all** task types — development and analysis / documentation / review alike — subject only to the safety switches in *Auto-clarity* and the *Boundaries* below (code, error text, destructive / security / ordered blocks stay in normal grammar). The task-type on/off split described in the rest of this section applies **only under `CAVEMAN=auto`**. |
| 8 | |
| 9 | Under **`CAVEMAN=auto`**, caveman is the default style for **development tasks**, where the user is acting as a senior 1C engineer and wants signal not prose: |
| 10 | |
| 11 | - writing or editing BSL / metadata XML / forms; |
| 12 | - refactoring; |
| 13 | - fixing bugs (including the systematic-debugging methodology); |
| 14 | - running shell commands, deploying, loading from / to an infobase; |
| 15 | - triaging a syntax / lint error; |
| 16 | - short technical Q&A about a specific code path or platform call. |
| 17 | |
| 18 | Under `CAVEMAN=auto`, caveman is **off** for **analysis, documentation and review tasks**, where the user needs structured, connected, audit-ready prose: |
| 19 | |
| 20 | - writing or updating PRDs, technical specifications, OpenSpec `proposal.md` / `design.md` / `tasks.md`; |
| 21 | - writing or updating user-facing or admin documentation, codemaps, API references (anything owned by `1c-doc-writer`); |
| 22 | - code review, architecture review, rule / process review, audit reports (anything owned by `1c-code-reviewer`, `1c-arch-reviewer`, or asked of the parent agent in review form); |
| 23 | - summaries, explanations and "teach me how X works" requests longer than a couple of sentences; |
| 24 | - handoff documents (`handoff` skill output); |
| 25 | - answers to "why" / "compare options" / "what are the trade-offs" questions where causality must be spelled out. |
| 26 | |
| 27 | When in doubt (still under `CAVEMAN=auto`), look at the verbs in the request: **"write" / "fix" / "refactor" / "deploy" / "run"** → caveman on; **"review" / "analyse" / "design" / "explain" / "compare" / "document" / "summarise" / "audit"** → caveman off. |
| 28 | |
| 29 | Two scopes for changing the state: |
| 30 | |
| 31 | - **Session-only** (this chat, no file change): "caveman please" forces on; "stop caveman" / "normal mode" / "обычный режим" forces off; `/caveman lite|full|ultra` switches the level. A forced session state overrides everything below and holds until the next force or session end. |
| 32 | - **Persistent** (project-wide, edits `.dev.env` `CAVEMAN`): the `/caveman on|off|auto` slash command (`content/commands/caveman.md`). |
| 33 | |
| 34 | ## Configuration — `.dev.env` (`CAVEMAN`) |
| 35 | |
| 36 | Automatic activation is gated by the `CAVEMAN` parameter in `.dev.env` (canonical description — `content/rules/dev-standards-env.md → "CAVEMAN — caveman auto-activation"`; toggled by the `/caveman on|off|auto` slash command, `content/commands/caveman.md`): |
| 37 | |
| 38 | - **`CAVEMAN=on`** (default / empty / invalid) — caveman is active for **all** tasks, development and analysis / documentation / review alike. The task-type split in *Scope* does not apply; only the *Auto-clarity* and *Boundaries* safety switches do. |
| 39 | - **`CAVEMAN=auto`** — task-type auto-classification as described in *Scope*: on for development, off for analysis / documentation / review. |
| 40 | - **`CAVEMAN=off`** — auto-activation is disabled: caveman never turns on by itself on any task. It stays off until the user issues an explicit in-session force command. |
| 41 | |
| 42 | **Precedence:** an explicit session force ("caveman please" / "stop caveman" / "normal mode", or a `/caveman lite|full|ultra` level switch) always wins → then the persistent `CAVEMAN` value (`on` → all tasks, `auto` → by task type, `off` → no auto-on). Read `.dev.env` for this value only when it is actually available; if the file or key is absent, treat it as `on`. |
| 43 | |
| 44 | ## Persistence |
| 45 | |
| 46 | When active (by default under `CAVEMAN=on`, by task-type classification under `CAVEMAN=auto`, or by force), caveman is ACTIVE FOR EVERY SUBSEQUENT RESPONSE within the same task. No filler drift. Under `CAVEMAN=auto` only: if the task shape changes (the user pivots from "fix this bug" to "write a PRD for the next feature"), re-classify and switch off accordingly. |
| 47 | |
| 48 | Default level when active: **full**. Switch with `/caveman lite`, `/caveman full`, or `/caveman ultra`. Level holds until session end or another swit |