Always-on ADHD-friendly output shaping: one system-prompt section makes replies action-first with numbered steps and concrete time estimates; adhd_on / adhd_off / adhd_status tools persist the mode across restarts.
Install
# from npm (prebuilt)
dsh plugin --profile web add dsh-i-have-adhd
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:yongshuai0314/dsh-i-have-adhd
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
ADHD-friendly output shaping for DeepSeek Harness (DSH): one system-prompt section that rewrites how the assistant replies — action first, numbered steps, concrete time estimates, zero preamble or closers — with live on/off switches that persist across restarts.
Inspired by ayghri/i-have-adhd (MIT). The ruleset here is an original rewrite for DSH; see THIRD-PARTY-NOTICES.md.
What changes
| Before | After |
|---|---|
| "Great question! Let me think about this. Your auth flow has a few moving pieces: the middleware, the token verification, and... Hope this helps!" | "Run npm install jsonwebtoken@latest, then edit src/auth.ts:42. 1. Open the file 2. Replace verifyToken 3. Run the tests. Next: paste the first failing line." |
How it works
- One system-prompt section (
dsh-i-have-adhd, order 50 — after the persona, before tool guidance) carries the ruleset into every model step while the mode is on. - Three zero-argument agent tools control it:
adhd_on— enable for this session and persist (flag file under$DSH_HOME/dsh-i-have-adhd/)adhd_off— disable and clear the flagadhd_status— report state, since when, and whether restart will restore it
- The persisted flag restores the mode at boot.
- The section registers through the systemPrompt service, so toggling takes effect on the very next model step — no session restart, no page refresh.
Usage
Install, then just say it in conversation:
adhd mode on → the model calls adhd_on; replies switch shape immediately
adhd mode off → back to default style, flag cleared
You can also ask for status any time, or create/remove the flag file yourself:
$DSH_HOME/dsh-i-have-adhd/always-on (default ~/.dsh/dsh-i-have-adhd/always-on).
The ruleset
Ten ideas, grouped for skimming — full text in dsh/rules.js:
Shape: first line is the action · number the steps · close with one next move · park side quests State: re-say where we are · estimate in real units · surface what now works Tone: errors are facts · cap lists at five · no throat-clearing
Explicit overrides: harness rules outrank the mode; "explain" gets full depth with the same shape; destructive actions still confirm; a three-turn debug spiral triggers one diagnostic question instead of more edits.
How this differs from other ADHD plugins
Two ADHD-related plugins already exist on the market, doing different jobs: dsh-adhd-copilot coaches the user (task breakdown, launch rituals); adhdgofly-dsh-ext highlights parts of speech in rendered Markdown. Neither touches how the assistant writes. This plugin shapes the assistant's own replies — action-first structure, live on/off, restart persistence.
Install
dsh plugin --profile <your-profile> add dsh-i-have-adhd
License
MIT. Third-party notices: the concept and name derive from ayghri/i-have-adhd (MIT), whose license text is preserved in THIRD-PARTY-NOTICES.md.
Links
More in this category
zhu1090093659/dsh-web#packages/dsh-skill-explorer★ 8597
Skill center for the dsh web GUI: browse all loaded skills grouped by source, enable or disable model invocation, create new skills, and delete into a recoverable trash.
GanyuanRan/Aegis★ 1335
Software-engineering method pack for coding agents, with skills for baseline-first planning, systematic debugging, prompt hygiene, verification before completion, and repair/retirement tracking.
superdesigndev/superdesign-skill★ 643
Design skill for UI and marketing graphics on the Superdesign canvas: reads the repo for context, extracts its design system, then generates and iterates branchable design drafts, flow pages, and reusable components through the Superdesign CLI.
linhay/harmony-next.skills★ 361
HarmonyOS NEXT skill bundle for DeepSeek Harness with offline API references and DevEco, HDC, and emulator automation guidance.
dhicoc/dsh-reverse-skill★ 235
Complete reverse-skill pack (85 SKILL.md) as a DeepSeek Harness Cordis plugin: reverse engineering, authorized pentesting and security-research skill router.
sandbaseai/sandbase-skills★ 203
Mounts 88 packaged research, social-intelligence, marketing and business Agent Skills into dsh through the filesystem Skill provider.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.