Skip to content
CAI
Software that uses CAICheck a score

ayghri/i-have-adhd

55.6

Adequate · 15 November 2026

1.6k

lines of production code

Python

with TypeScript

5

measurements over time

CAI band scale
CAI trend line
CAI lens gauges

What this system is

This system is a configuration tool and evaluation framework designed to modify AI assistant behavior for users with ADHD. It implements persistent output formatting rules that enforce structured, action-oriented responses while suppressing tangents. Additionally, it provides a structured harness to systematically evaluate and score these behavioral changes against baselines using blind-judge metrics.

Features

Add 'i-have-adhd' plugin for focused output formatting

The .opencode directory now includes a new plugin that provides a slash command and an optional always-on mode to shape AI responses for users with ADHD. This feature encourages leading with next actions, numbering multi-step work, suppressing tangents, and providing concrete time estimates. Users can activate this behavior via the \/i-have-adhd\ command or by enabling an always-on flag, which persists across turns until disabled.

.opencode · high confidence

Add ADHD mode extension with session-persistent toggle and config options

Introduces a new 'i-have-adhd' extension that injects a specific ruleset into the conversation context to modify the AI's output style. The extension supports a command-line flag (\--adhd\) and a configuration file (\i-have-adhd.json\) with \alwaysOn\ and \hideStatus\ options to control default behavior and UI visibility. It maintains session state by persisting the enabled/disabled status across conversation turns and handles context synchronization to ensure rules are re-injected if dropped by compaction. A compatibility layer (\context-compat.ts\) is added to safely inspect session context messages, failing open if the underlying API changes or errors occur.

extensions · high confidence

Add opt-in always-on ADHD mode via SessionStart hook

Users can now enable a persistent 'ADHD mode' that automatically injects the i-have-adhd ruleset into every session. To activate, create the flag file \.i-have-adhd-always\ in the Claude config directory (default \\~/.claude\). When active, the hook runs on session start (startup, resume, clear, or compact) to output the ruleset body, stripping any YAML frontmatter from the skill file. The feature is implemented as a Node.js script (\always-on.mjs\) with POSIX shell (\always-on.sh\) and PowerShell (\always-on.ps1\) fallbacks for cross-platform compatibility, and is configured via the new \hooks/hooks.json\ with a 30-second timeout to ensure it never blocks session initialization.

hooks · high confidence

Introduce structured evaluation harness for comparing response quality

A new evaluation harness is available in the \evals/\ directory to systematically compare response quality between different conditions (e.g., baseline vs. skill-injected). The harness includes a set of test cases (\cases.jsonl\), a scoring rubric (\rubric.md\), and example runner configurations (\runners.example.json\) that isolate the model from user-specific settings to ensure fair comparison. Users can run evaluations using \scripts/run\_evals.py\ to generate responses, then use \scripts/judge.py\ to score them using a blind-judge approach that permutes condition labels to prevent position bias. The system reports token usage, cost deltas, and quality scores across dimensions like correctness, autonomy, and safety, with a first-run result document (\RESULTS.md\) showing the candidate skill's performance against the baseline.

evals · high confidence

New evaluation and compatibility validation scripts

Added four new scripts to the \scripts/\ directory to support evaluation workflows and compatibility checks. \check\_context\_compat.ts\ validates that the Pi and OMP session-context APIs return the expected message structures and handle failures gracefully. \check\_pi\_extension.py\ performs smoke tests on the Pi/OMP package manifest and RPC interface without making model requests. \judge.py\ implements a blind-judge evaluation harness that scores paired responses across dimensions like correctness and safety while hiding condition identities. \run\_evals.py\ and \run\_scenario\_eval.py\ provide the underlying infrastructure to validate, run, and score these evaluations, including multi-turn scenario capture and cost tracking.

scripts · high confidence

Test coverage

29 commits adding/updating tests in tests; Add scenario to evaluate persistence through topic switches and stop commands.

Dependencies

Initial release of i-have-adhd package (v0.4.1)

Added the 'i-have-adhd' package, a new dependency that provides extensions and skills to shape output for users with ADHD. The package, version 0.4.1, is configured to load the 'i-have-adhd.ts' extension and includes skills from the './skills' directory, aiming to lead with next actions, number multi-step work, and suppress tangents.

(dependencies) · high confidence

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

How this codebase got here

Score

  • CAI 54 → 56 (+1.8)
  • Rubric changed (rubric-2026.11.1 → rubric-2026.11.4) — scores are not directly comparable.

Lenses

  • Code Health 49 → 51 (+2.2)
  • Architecture 98 → 98 (-0.4)
  • Maturity 62 → 66 (+3.7)
  • Readiness 41 → 43 (+2.0)
  • Security 95 → 92 (-2.7)
  • Performance 100 → 100 (+0.0)

Resolved (9)

  • Scanner failed to run — not a clean result
  • Scanner failed to run — not a clean result
  • judge.main (cognitive 21) (scripts/judge.py)
  • judge.main (cyclomatic 16) (scripts/judge.py)
  • run_evals.main (cognitive 23) (scripts/run_evals.py)
  • run_evals.run_evaluations (cognitive 57) (scripts/run_evals.py)
  • run_evals.run_evaluations (cyclomatic 30) (scripts/run_evals.py)
  • run_evals.summarize_usage (cognitive 43) (scripts/run_evals.py)
  • run_evals.summarize_usage (cyclomatic 29) (scripts/run_evals.py)

Written by cca.canine.dev from the codebase's own history, inside the signed delivery this page is composed from.

Survey your own repository

ayghri/i-have-adhd was measured the same way every project in this corpus was: the same rubric, at a pinned commit, with the result published in full. Point a surveyor at a repository you know and see whether you agree with it.

About this page

  • The score is its most recent published measurement, taken on 15 November 2026 at a pinned commit. It is not a live figure and does not change until the project is measured again.
  • Measured at commit 1ecea1993b7fce257ffbd686c070948b5e07eda6 — the exact code this score is about.
  • Scored under rubric-2026.11.4 — the same rubric and the same method as every other entry in this index.
  • Measured by Canine Code Analyzer (cca.canine.dev) using cca measure-2026.11.73+preprod-52de90b74216.