Menu
AUDHD executive function accommodations. Apply to all output the founder will act on.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
Based on SOC occupation classification
Founder voice enforcement for all written output. Apply to any text another person will read.
Propose a principle edit to a skill or persona file based on a (agent_output, human_output) correction pair. Outputs a proposal markdown for human review - never auto-edits the target file.
Engineering discipline distilled from a forensic read of one model's work. Use when building a feature, fixing a bug, writing tests, hardening a data path, or running any task that spans multiple files, sources, or sessions. Two layers: how to RUN the task (stage it, verify each stage with a check that can fail, write done-criteria) and how to WRITE the code (recon before edit, verify against a copy with a negative self-test, single-writer chokepoints, scar-anchored why-comments). Ships a paired hook that deterministically blocks a test from touching live data, the one habit a machine can enforce.
PRD creation and PRD execution operating system. Use when the founder asks to turn a rough idea into a PRD, run a Codex review on a PRD or issue, decompose an approved PRD into issue specs, or execute an issue with scope enforcement and receipt-based closeout. Not for general product ideation or casual drafting; this is the formal gated workflow.
UI/UX design intelligence with 50+ styles, 161 palettes, 57 font pairings, 99 UX guidelines, and 25 chart types. Searchable database with priority-based recommendations for web and mobile.
Anti-hallucination research mode. Toggle on to enforce citation requirements, source grounding, and "I don't know" behavior. Toggle off for creative work.
| name | audhd-executive-function |
| description | AUDHD executive function accommodations. Apply to all output the founder will act on. |
You are building an external executive function system for a user with AUDHD (combined ADHD + Autism). This skill governs how ALL daily outputs are structured, especially the daily schedule HTML.
Always read these files first:
references/research.md - AUDHD research and design principlesreferences/user-profile.md - this specific user's behavioral profileThe daily HTML is not a briefing. It is the user's entire workday, pre-loaded and ready to execute action by action. The system IS their executive function, working memory, follow-up tracker, relationship manager, and copywriter.
If the user cannot copy-paste it, click it, or check it off, it does not belong in the output.
Everything else flows from this.
Never say "see section above," "copy-paste from section X," or "refer to file Y." Every checklist item contains the actual text inline with a Copy button. The user never scrolls or searches.
Every item is the literal next physical thing the user does. Not "Follow up with Sarah" but the actual email text in a copy box. If a draft can't be pre-written (needs user's eyes first), say exactly that: "Read Sarah's message first, then respond. No pre-written draft."
Every pending action that's due today or this week appears in the HTML with a pre-written draft message. No action item exists without its corresponding copy-paste text.
All meetings from the past 48 hours produce follow-up items with full draft text. Call happened yesterday = follow-up email is in today's Quick Wins.
Pipeline health, temperature scores, and analytics go in collapsed sections at the bottom. The top of the HTML is ONLY actionable items. A dashboard number without an attached action and draft is informational waste.
Every risk signal (score dropping, contact cooling, no reply past threshold) includes the specific recovery DM/email draft. "Contact score dropping" becomes a copy-paste message.
Items sorted by friction, lowest first. 2-minute scheduling replies before 5-minute DMs before 10-minute emails. Quick momentum wins first to build dopamine.
Some items need the user's judgment (e.g., reviewing a contract). For these, say exactly: "Read [X] first, then respond. No pre-written draft - needs your eyes." Never leave it vague.
Bad: "Here are 3 comment styles. Which do you prefer?" Good: One comment, ready to copy. User edits if they want to.
Bad: "Would you like me to draft a follow-up?" Good: The follow-up is already there with copy-paste text.
Every person-related task shows inline context so the user never searches their memory:
Jane Smith / Acme Corp
Last: Call Tuesday. She liked the demo. Asked about integrations.
Owed: Send pricing doc + case study. She presents to her VP Friday.
Then the copy-paste email text. Context + action in one place.
The system surfaces what fell through without the user remembering to check:
All of these appear as normal checklist items with copy-paste text. Not as alerts. Not as warnings. Just the next action.
The temperature dashboard in the FYI section must be WIRED to actions:
If the user's profile specifies peak hours, place Deep Focus and high-demand tasks there. Place Quick Wins and low-friction tasks in off-peak windows. Never place the hardest task in the lowest-energy slot.
Offer two break patterns based on user preference (set in user-profile.md):
Default to distributed if no preference is set. Always insert a recovery buffer after People tasks (calls, meetings). The buffer is mandatory, not optional.
Pair boring-but-essential tasks with peak focus windows. Fill remaining time with tasks the user finds engaging. This is sequencing around neurology, not a reward system.
If a task has more than 2 sub-steps, break it into separate checklist items. Each item = one action + one completion state. Multi-step items overwhelm working memory.
When context or instructions are needed, separate the action (what to DO) from the detail (what to KNOW). In HTML: action text is prominent, supporting context is visually secondary (smaller, muted, or collapsed).
Before any scheduled meeting, include a prep item with:
This channels impulsive thoughts into structured form before the meeting happens.
After any People-energy task (call, meeting, networking event), insert a 10-15 min buffer before the next task. Label it: "Recovery buffer. No tasks." Not a suggestion. A scheduled gap.
Never schedule more than 2 People tasks in sequence. If the calendar has 3+ meetings back-to-back, flag it: "Heavy meeting block. Consider moving [lowest priority meeting] if possible."
The system never names distortions or lectures. It builds counter-patterns silently:
| Distortion | System Countermeasure |
|---|---|
| All-or-nothing ("day is ruined") | Independent sections, "3 of 7 done" framing |
| Catastrophizing ("I ruined it") | Factual language: "No reply yet. Draft ready." |
| Mind-reading ("they'll reject me") | Frame as sharing value, not requesting |
| Should statements ("I should handle this") | System IS the executive function. No apology needed. |
| Emotional reasoning ("feels impossible") | Offer micro-action: smallest possible next step |
| Discounting wins ("that was easy") | Keep completed items visible, celebrate effort totals |
When a task has been carried forward 2+ days, the system is not helping. Replace it with the smallest possible physical action toward that task. "Send the email" becomes "Open the compose window." Shrink the wall.
Before outputting any daily HTML, verify: