v0.2.0 — use it in claude.ai, ChatGPT, Gemini and more

Use it outside Claude Code

• claude.ai / Claude Desktop: download adhd-study-coach-skill.zip, then Settings → Capabilities → Skills → Upload skill.
• ChatGPT, Gemini, Copilot, other chat AIs, API/local models: paste a prompt.
  • adhd-study-coach-compact.md (~5k chars): start here; fits Custom GPT instruction limits.
  • adhd-study-coach-prompt.md (~10k chars): full version for Projects, Gems, API system prompts.
• Setup per provider: https://github.com/itsAalaa7/adhd-study-coach/blob/main/universal/README.md

Memory outside Claude Code: the coach prints a small Progress Card (JSON) after each session. Save it and paste it at the start of next time. Only the Claude Code plugin saves progress automatically.

Not yet tested on every provider; reports welcome via Issues.

Also in this release

• Plugin icon (.claude-plugin/icon.svg)
• Claude Code plugin behavior unchanged from v0.1.1

Not a clinical tool.

v0.1.1 — safety hardening + citation corrections

Citation corrections (RESEARCH.md)

A verification pass against publisher, PubMed and index pages found and fixed several errors:

• Adesope et al. (2017): the earlier "217 studies, g = 0.61" could not be confirmed. Now: 272 effect sizes; practice testing vs restudy g ≈ 0.51, vs no activity g ≈ 0.93 (from secondary sources; check the paper's abstract before quoting exact numbers).
• Rohrer & Taylor (2007): the earlier "roughly doubled next-day scores" was wrong. The test was one week later, and mixed practice scored far higher.
• Alderson et al. (2013, adults): softened "large effect sizes" to "significant deficits" (adult effect sizes not re-checked).
• Body doubling: removed an untraceable quote; evidence is anecdotal and recent studies are mixed.
• Rejection Sensitive Dysphoria: removed the unverified "~5 studies" count.
• Confirmed as accurate: Roediger & Karpicke 2006, Cepeda et al. 2006, Dunlosky et al. 2013, Barkley 1997, Kasper et al. 2012, Sonuga-Barke 2003, the 2025 gamification RCT, Brunmair & Richter 2019.
• Not re-checked: Sweller 1988, Bjork & Bjork, CHADD/NICE guidance, "Wall of Awful", Dodson on RSD.

See the "Verification status" table at the bottom of RESEARCH.md.

Other changes

• Agent tool allowlist (Read, Write, Edit, Glob, Grep, Bash, PowerShell) and explicit safety boundaries: data folder only, no network, stored notes treated as data.
• Visible progress line after warm-up; optional body-doubling offer.
• README rewrite (ADHD-friendly layout, safety table, example session); SECURITY.md added.

Not a clinical tool.