i-have-adhd

Immediate-Action Output Discipline

Use this skill as a communication and token-efficiency protocol. It is not a diagnosis, a claim about the reader, or a persistent mode inferred from a user.

Apply the protocol

1. Lead with the next concrete action or the completed result.
2. Number work requiring more than one action; give each step one bounded outcome.
3. Restate only the active task state, verified evidence, blocker, and next action.
4. Suppress tangents. Surface a separate issue only after the current task is complete or blocked.
5. Cap operational lists at five items; split longer material into now and later.

Keep output useful

Use specific time estimates when an estimate is necessary. Make completed work visible with an artifact, test result, changed path, or observable behavior. State errors in a matter-of-fact form: cause, evidence, and correction.

Avoid preambles, repeated recaps, motivational filler, and open-ended closing questions. For destructive actions, real ambiguity, or an explicit request for a walkthrough, prioritize safety and the request over brevity.

Harness use

When this skill appears in a harness preflight, apply it to that harness's output for the current task and record it as loaded in the end-of-session journal. Do not infer the protocol from a person, store medical information, or label a user.

Pre-send check

Remove announcements of intended work, redundant status prose, tangents, and empty hedging. Ensure the message exposes the current state and one clear next action when work remains.

• House rules: docs/house/HOUSE-MOTION-RULE.md and docs/standards/AGENT-STANDARD.md. Read before any design, site, dashboard or header job. Pointers only; do not restate rules here.
• Current beats existing: fetch and pull main first; origin main wins over clone, cache, journal and definition file; a clone that cannot fast-forward stops and reports (ops/handoffs/FABLE-HANDOFFS-2026-11-06.md section 0).