Caveman
JuliusBrussee/cavemanA token-saving pack that makes agents answer in terse "caveman" prose, plus an optional CLI with output trimming, model routing and usage telemetry.
With caveats
Broad access Depth 3/5
i-have-adhd Claude skills
A single output-style skill that makes a coding agent lead with the next action, number steps, drop preamble and keep lists short.
Our verdict Recommended
A small, clean MIT-licensed style skill that does what it says; it only changes reply formatting and the scan found nothing.
What i-have-adhd does
i-have-adhd changes how a coding agent writes its replies, not what it can do. After you invoke /i-have-adhd, the agent follows ten rules for every later response until you say 'stop adhd mode' or 'normal mode': lead with the next action, number multi-step tasks, end with one concrete next step, suppress tangents, restate the current state each turn, give time estimates in specific minutes, make completed steps visible, state errors matter-of-factly, cap lists at five items, and skip preamble, recaps and closing pleasantries. The SKILL.md (about 7 KB) explains the reasoning for each rule, for example that working memory is small and that vague time estimates feel uniform, and gives bad and good examples such as replacing a paragraph about an auth flow with 'Edit src/auth.ts:42' plus numbered steps.
There are no scripts or references. The repo ships the same skill for several agents: a Claude Code plugin manifest, a Cursor copy, and agent files for OpenAI/Codex and Gemini, plus a few eval files and translated READMEs. The README says it is loosely based on a book on adult ADHD and that no diagnosis is needed to use it. You can fork and edit the SKILL.md to tune the rules.
Good for
- Developers who find long, chatty agent replies hard to act on
- Making agent output action-first and scannable for anyone, not only people with ADHD
- A tiny, auditable style skill you can fork and adjust
Not for
- Tasks where you want the agent's full reasoning and context in the reply
- Teams that need a documented, tested workflow rather than a formatting style
What's inside depth 2/5
One 7 KB SKILL.md of rules with examples and a handful of eval files, but no references or scripts.
- Skills
- 2
- Reference docs
- 0
- Scripts
- 0
- Tests / evals
- 7
- Templates
- 0
- Avg SKILL.md
- 7 KB
Counted from the repository file tree. How we score depth.
Security review
Review before installingA pure instruction skill (SKILL.md with disable-model-invocation) that changes response style; it has no network calls of its own. Plugin packaging adds a SessionStart hook and plugin/extension code for several agents, which only inject the ruleset when you create an opt-in flag file. The scripts/ folder holds developer eval harnesses that shell out to claude/codex CLIs and are not used by the skill; install docs use unpinned npx skills and main-branch downloads.
Worth knowing
- Changes files outside your projectOptional install instructions have the user create dotfiles/commands in agent config folders (e.g. curl the vendor's gemini.toml into ~/.gemini/commands, touch ~/.claude/.i-have-adhd-always); nothing in the repo code does this automatically.INSTALL.md:310
- Makes network requestsGemini install route downloads a command file from raw.githubusercontent.com/ayghri/i-have-adhd/main (vendor's own repo, unpinned branch) to disk; the skill itself does not fetch or follow remote text at runtime.INSTALL.md:310
- Runs remote code at an unpinned versionInstall docs use
npx skills add ayghri/i-have-adhd(npm 'skills' CLI, no version pin) and marketplace/extension installs that track the main branch.INSTALL.md:357 - Relies on a CLI or package we could not reviewInstall/update docs rely on third-party CLIs (npx skills, agy, grok, gemini, omp, qwen) whose source is not in this repo.INSTALL.md:357
What we checked
The repository (74 files) on Oct 29, 2026. Not a full review: we read 19 of 27 code files; it relies on code outside the repository: npm package 'skills' CLI (run by `npx skills add/update/remove` in the install docs); Agent CLIs used for installation and by the dev eval scripts (claude, codex, agy, grok, gemini, omp, pi, qwen); @earendil-works/pi-coding-agent (imported by extensions/i-have-adhd.ts).
NVIDIA SkillSpector static scan: 490 pattern hits (MCP Rug Pull 169, Prompt Injection 108, Agent Snooping 97, Rogue Agent 60, Tool Misuse 32, Dangerous Code Execution 11). SkillSpector hits are documentation mentions (rm of the opt-in flag file, mkdir of agent config dirs, 'npx skills', zero-width joiners in translated Persian docs), a rubric comment about the eval grader, and os.environ copies in dev scripts/tests that strip API keys before running a smoke test. No hidden instructions or obfuscated code found.
A review, not a certification. Skills change after we check them — read the files before installing.
License
MIT — commercial use allowed.
Can it run unattended?
Not really. It is a response-formatting style for interactive use and has no loop, stop conditions or verification steps.
Alternatives to i-have-adhd
Ponytail
DietrichGebert/ponytailA "lazy senior dev" skill pack that pushes coding agents toward the smallest complete change, with review, audit and cost-tracking commands.
Recommended
Check first Depth 2/5
Matt Pocock's Skills
mattpocock/skillsSmall, composable engineering skills: grill-me interviews, shared glossary, TDD, triage, prototyping, code review, handoff and a flow router.
Recommended
Check first Depth 2/5
FAQ
What is the i-have-adhd skill?
It is an agent skill that changes the shape of a coding agent's replies: next action first, numbered steps, one concrete next step, short lists, specific time estimates and no filler. You switch it on with /i-have-adhd and off with 'stop adhd mode'.
How do I install i-have-adhd?
The README suggests telling your agent to install the skill from github.com/ayghri/i-have-adhd and follow the repo's AGENTS.md, or you can follow INSTALL.md. For Claude Code the repo is also a plugin marketplace you can add and install from.
Do I need an ADHD diagnosis to use it?
No. The README says no diagnosis is needed; the rules just produce shorter, action-first replies that many people find easier to follow.
i-have-adhd vs caveman: what is the difference?
i-have-adhd restructures replies around next actions and numbered steps for readability. Caveman is aimed at cutting tokens with very terse output. They overlap in brevity but have different goals.
Sources: GitHub repository (stars, license, file tree, last push Oct 19, 2026). Review based on the README and SKILL.md files as published.