A skill to stop your coding agent from burying the answer. ADHD-friendly output.
How to install
Antigravity (agy)
Install
agy plugin install https://github.com/ayghri/i-have-adhd
Verify
agy plugin list
Update
agy plugin uninstall i-have-adhd
agy plugin install https://github.com/ayghri/i-have-adhd
Uninstall
agy plugin uninstall i-have-adhd
Or keep it installed and turn it off: agy plugin disable i-have-adhd.
Always-on (optional)
Add to ~/.gemini/GEMINI.md:
## Output style
The reader has ADHD. Shape every response so it can be acted on:
1. Lead with the answer or next action: command, path, or snippet first.
2. Number multi-step work; one bounded action per step.
3. End with one next action doable in under two minutes.
4. Finish the current issue before raising a new one.
5. Restate progress each turn ("step 3 of 5 done").
6. Give time estimates in concrete units, never "a bit".
7. After a change, show what now works.
8. Errors: state location, cause, and fix. No drama.
9. Cap lists to 5 items.
10. No preamble, no recaps, no closers.
Exceptions: explain fully when asked to explain. Confirm before destructive actions. After three failed fixes, stop and name the doubtful assumption. If the request is ambiguous, ask one short question.
AstronClaw (custom skill)
AstronClaw supports importing a Markdown file as a custom skill. This route uses
the existing SKILL.md; see its official skills guide
for the upload and management controls.
This procedure follows AstronClaw's documentation but has not been tested with this skill. Check the exported instructions before enabling it.
Install
- Download the canonical SKILL.md and save it as
SKILL.md. Review its contents before uploading. - In AstronClaw, open 我的技能 (My skills), choose 新建 (New), and upload that
.mdfile. - Check that the imported skill is named
i-have-adhd. Use 启用/禁用 (Enable/Disable) to control its availability.
Only the skill Markdown is needed. Uploading sends that file to AstronClaw; the repository's plugin manifests and hooks are not part of this setup.
Verify and activate
Confirm i-have-adhd appears in My skills. Use 下载 (Download) to review
the imported instructions against the original skill, then enable it and try:
Use the i-have-adhd skill for this conversation. Explain how to create an empty Git repository in a new folder.
Check that the reply leads with the action and numbers the steps. This is a manual check of the imported skill; a successful upload alone does not verify that its response rules are being applied.
Activation note
AstronClaw supports both explicit requests and automatic skill invocation.
Its guide does not specify whether it honors disable-model-invocation: true,
so use Disable when you do not want the skill available. There is no need
to rely on a /i-have-adhd slash command.
The skill instructs the assistant to keep the style for the conversation until
you say stop adhd mode or normal mode. That instruction does not change the
platform toggle; disable the skill and start a new conversation for a fresh
session without it.
Update
Download the latest canonical SKILL.md. If you customized the imported copy,
use 下载 (Download) to keep a backup first. For a clean replacement, delete
the old i-have-adhd entry, repeat the import, and run the verification prompt
in a new conversation.
Uninstall
In My skills, select i-have-adhd and choose 删除 (Delete), then start a
new conversation. To keep the imported copy for later, choose Disable instead.
Claude Code
Install
claude plugin marketplace add ayghri/i-have-adhd
claude plugin install i-have-adhd@i-have-adhd
Type /i-have-adhd.
Verify
claude plugin list
Update
claude plugin marketplace update i-have-adhd
claude plugin update i-have-adhd@i-have-adhd
The first command refreshes the marketplace listing. The second updates the installed plugin. Restart Claude Code to load it.
claude plugin update skips a plugin whose version has not changed. If you maintain a fork, bump version in .claude-plugin/plugin.json after you edit the rules.
Uninstall
claude plugin uninstall i-have-adhd
claude plugin marketplace remove i-have-adhd
Or keep it installed and turn it off: claude plugin disable i-have-adhd.
Always-on (optional)
A SessionStart hook loads the full ruleset at the start of every session, no /i-have-adhd needed:
touch ~/.claude/.i-have-adhd-always
If you use a custom Claude configuration directory, create the flag there instead:
touch "$CLAUDE_CONFIG_DIR/.i-have-adhd-always"
Back to on-demand:
rm ~/.claude/.i-have-adhd-always
The hook only fires when the flag file exists, so installing the plugin changes nothing by itself. "stop adhd mode" still turns it off for the current session.
Codex
Install
codex plugin marketplace add ayghri/i-have-adhd --ref main
codex plugin add i-have-adhd@i-have-adhd
Invoke the skill explicitly by typing $i-have-adhd. Codex will not activate
it automatically.
Verify
codex plugin list
Update
codex plugin marketplace upgrade i-have-adhd
codex plugin remove i-have-adhd
codex plugin add i-have-adhd@i-have-adhd
Uninstall
codex plugin remove i-have-adhd
codex plugin marketplace remove i-have-adhd
Always-on (optional)
Add to ~/.codex/AGENTS.md:
## Output style
The reader has ADHD. Shape every response so it can be acted on:
1. Lead with the answer or next action: command, path, or snippet first.
2. Number multi-step work; one bounded action per step.
3. End with one next action doable in under two minutes.
4. Finish the current issue before raising a new one.
5. Restate progress each turn ("step 3 of 5 done").
6. Give time estimates in concrete units, never "a bit".
7. After a change, show what now works.
8. Errors: state location, cause, and fix. No drama.
9. Cap lists to 5 items.
10. No preamble, no recaps, no closers.
Exceptions: explain fully when asked to explain. Confirm before destructive actions. After three failed fixes, stop and name the doubtful assumption. If the request is ambiguous, ask one short question.
Grok (grok)
Grok loads the repository's existing plugin and skill files; no separate Grok manifest is required. Install directly from GitHub, enable the plugin, then invoke the skill. Two Grok-only steps: --trust (hooks and skills stay inactive without it) and grok plugin enable (plugins stay off until enabled).
Install
grok plugin install ayghri/i-have-adhd --trust
grok plugin enable i-have-adhd
Start a new Grok session and type /i-have-adhd. Grok honors disable-model-invocation: true, so nothing applies until you invoke the skill or turn on always-on.
Verify
grok plugin list
grok plugin details i-have-adhd
Confirm i-have-adhd is listed, enabled, and shows a skill plus hooks.
Update
grok plugin update i-have-adhd
Uninstall
grok plugin uninstall i-have-adhd --confirm
Or keep it installed and turn it off: grok plugin disable i-have-adhd.
Always-on (optional)
Add the block to ~/.grok/AGENTS.md, or drop it in ~/.grok/rules/i-have-adhd.md (Grok loads both at session start):
## Output style
The reader has ADHD. Shape every response so it can be acted on:
1. Lead with the answer or next action: command, path, or snippet first.
2. Number multi-step work; one bounded action per step.
3. End with one next action doable in under two minutes.
4. Finish the current issue before raising a new one.
5. Restate progress each turn ("step 3 of 5 done").
6. Give time estimates in concrete units, never "a bit".
7. After a change, show what now works.
8. Errors: state location, cause, and fix. No drama.
9. Cap lists to 5 items.
10. No preamble, no recaps, no closers.
Exceptions: explain fully when asked to explain. Confirm before destructive actions. After three failed fixes, stop and name the doubtful assumption. If the request is ambiguous, ask one short question.
Gemini CLI
Gemini CLI has no plugin marketplace, so there are two native routes: a custom command (opt-in, off until you invoke it) or an extension (always-on once installed). The command route matches this skill's default posture; pick it unless you want the rules on every session.
Install (command, opt-in)
mkdir -p ~/.gemini/commands
curl -fsSL https://raw.githubusercontent.com/ayghri/i-have-adhd/main/skills/i-have-adhd/agents/gemini.toml \
-o ~/.gemini/commands/i-have-adhd.toml
Start a new session, type /i-have-adhd. It stays on for that session.
Install (extension, always-on)
gemini extensions install https://github.com/ayghri/i-have-adhd
The extension loads GEMINI.md, which imports the full skill, so the rules apply from message one. git must be installed.
Verify
gemini extensions list # extension route
ls ~/.gemini/commands # command route: i-have-adhd.toml present
Or type / in a session and confirm i-have-adhd is listed.
Update
gemini extensions update i-have-adhd # extension route
# command route: re-run the curl above
Uninstall
gemini extensions uninstall i-have-adhd # extension route
rm ~/.gemini/commands/i-have-adhd.toml # command route
GitHub Copilot (VS Code and Copilot CLI)
Copilot reads Agent Skills natively: the same SKILL.md, no conversion. It scans .github/skills/, .claude/skills/, and .agents/skills/ in the project, and ~/.copilot/skills/, ~/.claude/skills/, and ~/.agents/skills/ globally.
Install
npx skills add ayghri/i-have-adhd -a github-copilot # this project
npx skills add ayghri/i-have-adhd -a github-copilot -g # all projects
Without the CLI, copy the skill folder into any directory Copilot scans:
git clone https://github.com/ayghri/i-have-adhd
mkdir -p ~/.copilot/skills
cp -R i-have-adhd/skills/i-have-adhd ~/.copilot/skills/
Verify
Type / in the chat input and confirm i-have-adhd appears. Or:
npx skills list
npx skills ls -g # if installed globally
Update
npx skills update i-have-adhd
Or re-copy the folder after git pull.
Uninstall
npx skills remove i-have-adhd
Or delete the i-have-adhd folder from the skills directory it landed in.
Activation note
Copilot respects disable-model-invocation: nothing applies until you invoke the skill, same as Claude Code (tested in #60).
Always-on (optional)
Add the block below to .github/copilot-instructions.md in the project (Copilot reads it into every chat):
## Output style
The reader has ADHD. Shape every response so it can be acted on:
1. Lead with the answer or next action: command, path, or snippet first.
2. Number multi-step work; one bounded action per step.
3. End with one next action doable in under two minutes.
4. Finish the current issue before raising a new one.
5. Restate progress each turn ("step 3 of 5 done").
6. Give time estimates in concrete units, never "a bit".
7. After a change, show what now works.
8. Errors: state location, cause, and fix. No drama.
9. Cap lists to 5 items.
10. No preamble, no recaps, no closers.
Exceptions: explain fully when asked to explain. Confirm before destructive actions. After three failed fixes, stop and name the doubtful assumption. If the request is ambiguous, ask one short question.
Hermes
Install
hermes skills install ayghri/i-have-adhd/skills/i-have-adhd
Type /i-have-adhd. The skill installs into ~/.hermes/skills/ and is exposed as a slash command at the next session start.
Prefer to browse first? Add this repo as a skill source (a "tap"), then search and install:
hermes skills tap add ayghri/i-have-adhd
hermes skills search adhd
hermes skills install ayghri/i-have-adhd/skills/i-have-adhd
Verify
hermes skills list
Update
hermes skills update i-have-adhd
Uninstall
hermes skills uninstall i-have-adhd
Or remove the tap too: hermes skills tap remove ayghri/i-have-adhd.
Always-on (optional)
Add to the AGENTS.md in your working directory (Hermes loads it per workdir), or to your persona SOUL.md for every session:
## Output style
The reader has ADHD. Shape every response so it can be acted on:
1. Lead with the answer or next action: command, path, or snippet first.
2. Number multi-step work; one bounded action per step.
3. End with one next action doable in under two minutes.
4. Finish the current issue before raising a new one.
5. Restate progress each turn ("step 3 of 5 done").
6. Give time estimates in concrete units, never "a bit".
7. After a change, show what now works.
8. Errors: state location, cause, and fix. No drama.
9. Cap lists to 5 items.
10. No preamble, no recaps, no closers.
Exceptions: explain fully when asked to explain. Confirm before destructive actions. After three failed fixes, stop and name the doubtful assumption. If the request is ambiguous, ask one short question.
Kimi Code CLI
Install
Start a Kimi Code session, then:
- Run
/plugins. - Choose Custom.
- Paste
https://github.com/ayghri/i-have-adhdand pressEnter. - Choose Trust and install.
Use slash command /skill:i-have-adhd to invoke the skill explicitly.
Update
/plugins in Kimi Code session, cursor to I Have ADHD, press R.
Uninstall
/plugins in Kimi Code session, cursor to I Have ADHD, press D.
OpenCode
The plugin supports OpenCode V1 1.18.29+ and V2. It registers the skill and /i-have-adhd command, and adds optional always-on mode. Older V1 versions need an upgrade or the plugin from the previous skill release.
Install
For use across projects:
git clone https://github.com/ayghri/i-have-adhd ~/.config/opencode/vendor/i-have-adhd
mkdir -p ~/.config/opencode/plugins
cat > ~/.config/opencode/plugins/i-have-adhd.js <<'EOF'
export { default } from '../vendor/i-have-adhd/.opencode/plugins/i-have-adhd.mjs';
EOF
For an existing V1 installation, keep the plugin entry in opencode.json. If switching to the loader above, remove that entry.
Start a new session and type /i-have-adhd. The rules apply until you say stop adhd mode or normal mode.
Skill and command only (without always-on)
Alternatively, from a checkout, copy the skill and command into OpenCode's global directories:
mkdir -p ~/.config/opencode/skills/i-have-adhd ~/.config/opencode/commands
cp skills/i-have-adhd/SKILL.md ~/.config/opencode/skills/i-have-adhd/SKILL.md
cp .opencode/command/i-have-adhd.md ~/.config/opencode/commands/i-have-adhd.md
Verify
Start OpenCode, type /, and confirm i-have-adhd appears in the command list.
Update
git -C ~/.config/opencode/vendor/i-have-adhd pull
For the copy-based install, pull your checkout and re-copy the two files.
Uninstall
Remove ~/.config/opencode/plugins/i-have-adhd.js. For the copy-based install, remove ~/.config/opencode/skills/i-have-adhd/ and ~/.config/opencode/commands/i-have-adhd.md. For an existing V1 config-based install, remove its plugin entry.
Always-on (optional)
Requires the plugin. The flag adds the complete ruleset to the system prompt on every turn:
touch ~/.config/opencode/.i-have-adhd-always
stop adhd mode or normal mode asks the agent to return to its normal style for the current session. Delete the flag to stop automatic injection:
rm ~/.config/opencode/.i-have-adhd-always
If XDG_CONFIG_HOME is set, replace ~/.config in these paths with that directory.
Pi
Pi discovers this repository as a native package: extensions/ provides the session-persistent mode and skills/ keeps the Agent Skills entry point available.
Install
pi install https://github.com/ayghri/i-have-adhd
Start a new Pi session. Toggle ADHD-friendly output for the current session:
/i-have-adhd
The footer shows ● ADHD ON while the mode is active. Run the command again to turn it off, or be explicit:
/i-have-adhd on
/i-have-adhd off
stop adhd mode
Like the Claude Code hook, the extension adds the ruleset to the conversation once instead of rewriting the system prompt on every request, and adds it again after compaction drops it.
The existing Agent Skills command remains available as an alias:
/skill:i-have-adhd
Start a new Pi session with the mode enabled by default:
pi --adhd
Verify
pi list
Confirm the GitHub package is listed, then type /i-have-adhd and check that ● ADHD ON appears in the footer.
Update
pi update https://github.com/ayghri/i-have-adhd
Or update every unpinned Pi package with pi update --extensions.
Uninstall
pi remove https://github.com/ayghri/i-have-adhd
Always-on (optional)
Create a flag in Pi's agent configuration directory:
touch ~/.pi/agent/.i-have-adhd-always
The extension checks the flag at every new, resumed, forked, or reloaded session. A saved choice for the current session wins over this default, so stop adhd mode keeps that session disabled.
Back to on-demand:
rm ~/.pi/agent/.i-have-adhd-always
Config file (optional)
Create ~/.pi/agent/i-have-adhd.json in Pi's agent configuration directory:
{
"alwaysOn": true,
"hideStatus": true
}
alwaysOn: start every session with the rules active — same as the.i-have-adhd-alwaysflag file, which still workshideStatus: keep the● ADHD ONstatus-bar entry hidden; the rules and the/i-have-adhdcommand still work
Read once at extension startup, so restart Pi after changing it. A saved choice for the current session wins over alwaysOn, so stop adhd mode keeps that session disabled.
If PI_CODING_AGENT_DIR is set, put .i-have-adhd-always in that directory instead. Run /reload or start a new session after changing the flag.
Oh My Pi (OMP)
Install
omp plugin marketplace add ayghri/i-have-adhd
omp plugin install --scope user i-have-adhd@i-have-adhd
Start a new OMP session and run /i-have-adhd to toggle the mode.
Update
omp plugin marketplace update i-have-adhd
omp plugin upgrade --scope user i-have-adhd@i-have-adhd
Uninstall
omp plugin uninstall --scope user i-have-adhd@i-have-adhd
omp plugin marketplace remove i-have-adhd
Qwen Code
Install
qwen extensions install ayghri/i-have-adhd
Qwen Code supports the GitHub shorthand and installs the repository as a
native extension. The extension discovers the skill under skills/.
Type /i-have-adhd to invoke the skill explicitly. Installing the extension
does not change output until the skill is invoked.
Verify
qwen extensions list
Then start a new Qwen Code session and run:
/skills
Confirm that i-have-adhd appears in the list.
Update
qwen extensions update i-have-adhd
Uninstall
qwen extensions uninstall i-have-adhd
Zed
Zed's Agent reads Agent Skills natively using the same SKILL.md format without conversion. Note that Zed's older "Rules" have been replaced by Skills alongside AGENTS.md instructions.
Install
In the Agent Panel, open the Skills manager and choose Create skill from URL (also in the command palette as agent: create skill from url), then paste:
https://github.com/ayghri/i-have-adhd/blob/main/skills/i-have-adhd/SKILL.md
Save it in User scope for every project, or Project scope for one. Then type /i-have-adhd in the Agent Panel.
Prefer the filesystem? Clone the repo and drop the skill folder into your user skills directory:
git clone https://github.com/ayghri/i-have-adhd
mkdir -p ~/.agents/skills
cp -R i-have-adhd/skills/i-have-adhd ~/.agents/skills/
Verify
Open the Skills manager in the Agent Panel and confirm i-have-adhd is listed. Or type / and confirm it appears.
Update
Re-import from the same URL (overwrites), or re-copy the folder after git pull.
Uninstall
Remove i-have-adhd from the Skills manager, or delete ~/.agents/skills/i-have-adhd.
Always-on (optional)
Add to your personal ~/.config/zed/AGENTS.md:
## Output style
The reader has ADHD. Shape every response so it can be acted on:
1. Lead with the answer or next action: command, path, or snippet first.
2. Number multi-step work; one bounded action per step.
3. End with one next action doable in under two minutes.
4. Finish the current issue before raising a new one.
5. Restate progress each turn ("step 3 of 5 done").
6. Give time estimates in concrete units, never "a bit".
7. After a change, show what now works.
8. Errors: state location, cause, and fix. No drama.
9. Cap lists to 5 items.
10. No preamble, no recaps, no closers.
Exceptions: explain fully when asked to explain. Confirm before destructive actions. After three failed fixes, stop and name the doubtful assumption. If the request is ambiguous, ask one short question.
Cursor, Amp, and any other agent-skills harness
Works with any harness that reads agent skills. Swap -a <agent> for yours.
Install
npx skills add ayghri/i-have-adhd # this workspace
npx skills add ayghri/i-have-adhd -g # all projects
npx skills add ayghri/i-have-adhd -a cursor -y # one agent only
npx skills add ayghri/i-have-adhd -a opencode -y
New agent chat, type /i-have-adhd.
Without the CLI, copy the skill folder into whatever path your agent scans:
git clone https://github.com/ayghri/i-have-adhd
mkdir -p ~/.cursor/skills # Cursor. Use .agents/skills for OpenCode, or your agent's own path
cp -R i-have-adhd/skills/i-have-adhd ~/.cursor/skills/
Verify
npx skills list
npx skills ls -g # if installed globally
Update
npx skills update i-have-adhd
npx skills update -g # if installed globally
Uninstall
npx skills remove i-have-adhd
npx skills remove i-have-adhd -g # if installed globally
Always-on (optional)
Paste this into your agent's persistent rules file. Cursor: Settings → Rules → User Rules, or a project rule under .cursor/rules/ with alwaysApply: true. OpenCode: ~/.config/opencode/AGENTS.md.
## Output style
The reader has ADHD. Shape every response so it can be acted on:
1. Lead with the answer or next action: command, path, or snippet first.
2. Number multi-step work; one bounded action per step.
3. End with one next action doable in under two minutes.
4. Finish the current issue before raising a new one.
5. Restate progress each turn ("step 3 of 5 done").
6. Give time estimates in concrete units, never "a bit".
7. After a change, show what now works.
8. Errors: state location, cause, and fix. No drama.
9. Cap lists to 5 items.
10. No preamble, no recaps, no closers.
Exceptions: explain fully when asked to explain. Confirm before destructive actions. After three failed fixes, stop and name the doubtful assumption. If the request is ambiguous, ask one short question.
How activation works
- Installed, not invoked. In Claude Code, Qwen Code, Codex, and Grok, nothing happens until you invoke the skill explicitly. Claude Code, Qwen Code, and Grok honor
disable-model-invocation: trueinSKILL.md; Codex honorspolicy.allow_implicit_invocation: falseinagents/openai.yaml. Other harnesses may load every skill's description at startup and activate the skill themselves. - You invoke it explicitly. Type
/i-have-adhdin Claude Code, Qwen Code, or Grok, or$i-have-adhdin Codex. Rules stay on for that session. "stop adhd mode" or "normal mode" turns them off. - You touch
~/.claude/.i-have-adhd-always(Claude Code). ASessionStarthook loads the full ruleset from message one, every session. - You add the always-on snippet above (Grok, Codex, and other harnesses). Grok reads
~/.grok/AGENTS.mdand~/.grok/rules/*.md. Keeps the core rules in your agent's persistent context.
In Claude Code, Qwen Code, Codex, and Grok, no middle ground: if you did not turn it on, it is off.
Troubleshooting
/i-have-adhd not in autocomplete. Restart the agent. The plugin index is read at startup. On Grok, also run grok plugin enable i-have-adhd and confirm the install used --trust.
Always-on flag has no effect. Update the plugin (claude plugin marketplace update i-have-adhd, then claude plugin update i-have-adhd@i-have-adhd) and restart. Hooks are read at startup, and the flag needs the plugin version that ships hooks/hooks.json. Grok does not read ~/.claude/.i-have-adhd-always; put the always-on block in ~/.grok/AGENTS.md or ~/.grok/rules/i-have-adhd.md.
claude plugin marketplace add fails. Use the owner/repo form. A local path must point at the repo root, not .claude-plugin/.
grok plugin install does nothing visible. Add --trust, then run grok plugin enable i-have-adhd, then start a new session. Grok plugins stay off and untrusted until those two steps.
Installed but replies still preamble. Open a new session. If it still drifts, tighten the wording in skills/i-have-adhd/SKILL.md.
Want different rules. Fork, edit skills/i-have-adhd/SKILL.md, then swap your copy in:
claude plugin uninstall i-have-adhd # drop the upstream copy first:
claude plugin marketplace remove i-have-adhd # fork and upstream share both names
claude plugin marketplace add <your-username>/i-have-adhd
claude plugin install i-have-adhd@i-have-adhd
Restart, then re-invoke /i-have-adhd.
Skill missing after npx skills add. Start a new agent chat. Skills are indexed at session start. Confirm the folder landed where your agent scans (~/.cursor/skills/ for Cursor, .agents/skills/ for OpenCode) and that the frontmatter name matches the folder name.