8000 A lot of overhead, couldn't this be an output-style instead of a skill? · Issue #187 · ayghri/i-have-adhd · GitHub
[go: up one dir, main page]
Skip to content

A lot of overhead, couldn't this be an output-style instead of a skill? #187

New issue
New issue
Open
Open
A lot of overhead, couldn't this be an output-style instead of a skill?#187

Description

@thevfxsupervisor
thevfxsupervisor
opened on Sep 11, 2026

I'm more than likely missing what some of the functionality is. I'm not sure I understand what all the scripts and tests do. But I wondered if this awesome skill would be better implemented as and output-style. I've made a similar attempt at concise output that way, btu your skill is more indepth for sure.
thanks for any input, newb vibe coder here

Activity

  1. brianjcohen commented on Sep 11, 2026

    @brianjcohen
    brianjcohen
    on Sep 11, 2026

    Output styles are Claude-only and I believe they were actually deprecated by Anthropic. Really the guts of this is just a skill and it can be used with most/any agentic harness.

  2. ayghri commented on Sep 11, 2026

    @ayghri
    ayghri
    on Sep 11, 2026
    Owner

    we're all vibing here in this repo and exploring to see what workflow is best for various tasks/depth.
    If you tested something that you think is better, feel free to share it here or open a PR and we can discuss it there.

  3. jpvajda commented on Sep 11, 2026

    @jpvajda
    jpvajda
    on Sep 11, 2026

    Really helpful to stay focused!

    Though I didn't feel like I needed everything in this repo as what actually matters: .cursor/skills/i-have-adhd/SKILL.md — that's the entire skill.

    How I installed it in ~30 seconds without installing anything:

    1. Copied that one file to a shared location: ~/.agents/skills/i-have-adhd/SKILL.md
    2. Symlinked it into each tool's skills folder: ~/.cursor/skills/i-have-adhd → ~/.claude/skills/i-have-adhd → same target
    3. Done — works in both Cursor and Claude, one source of truth, no per-tool duplication.

    To simplify things even further: I just set this as a User Rule in Cursor and as a instruction set in Claude so i don't need to invoke the skill manually.

    Claude: Settings → Profile → Instructions for Claude
    Cursor: Cursor Settings -> Rules -> User rule

    I also wasn't clear why this skill had so much additional code as I feel this is all you need, maybe I'm missing something?

    It works great as a rule BTW.

    Rule Definition

    The reader has ADHD. Shape every response so it can be acted on immediately, not just read. 
1. Lead with the next action - answer, command, path, or snippet first. Context/prose after, if at all. 
2. Number multi-step tasks - one bounded action per step, fewest steps possible. 
3. End with one concrete next action - something doable in under two minutes. 
4. Suppress tangents - finish the current topic; surface unrelated issues once at the end as a separate item. 
5. Restate state every turn on multi-step work (e.g. "Step 3 of 5 done: X. Next: Y"). 
6. Give specific time estimates in concrete units, never vague ("some work").
7. Make completed work visible in concrete terms, don't bury wins in a recap. 
8. Matter-of-fact tone for errors - state cause and fix, no "uh oh." 
9. Cap lists to 5 items, most relevant first; retain the rest internally. 
10. No preamble ("Great question," "Let me..."), no recap, no closing pleasantries ("let me know if you need anything else").  

Exceptions: explain fully when asked to "explain/walk me through" (still no preamble/closer); confirm before destructive actions; stop and ask one diagnostic question after 3 failed debug attempts; ask one clarifying question on real ambiguity; when a rule would delete the actual answer (e.g. "what are my options"), keep the content but keep the shape (ranked, concise, recommendation first).

  4. b1gb3n97 commented on Sep 15, 2026

    @b1gb3n97
    b1gb3n97
    on Sep 15, 2026

    Output styles are Claude-only and I believe they were actually deprecated by Anthropic.

    This is wrong. https://code.claude.com/docs/en/output-styles

    I don't see why this repo can't offer an output style for Claude Code users. It works much better because once set, an output style not only applies to every conversation, but every response. It cuts out all the extra song and dance required to re-inject or force the agent to adhere to a certain style. Anthropic prescribes this exact feature for this scenario.

    And you can easily switch them on/off using /config.

  5. sethlivingston commented on Sep 17, 2026

    @sethlivingston
    sethlivingston
    on Sep 17, 2026

    For those suggesting making this an output style: Claude rarely adheres to my output style instructions. Your mileage may vary.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions
      0