Skip to main content

ac-ideate

Run Auto-Claude AI-powered ideation for any project directory. Generates low-hanging fruit, high-value features, and UI/UX ideas using the Auto-Claude backend. Use when brainstorming new specs, features, or architectural directions for a project.

Informations de source

Dépôt
ryanmaclean/skills
Dernière activité de la source
15 avril 2026 à 03:58
Langue détectée de SKILL.md
anglais
Étoiles
1
Forks
0

Options d'installation

Le prompt qui vérifie d'abord la source est sélectionné par défaut. Vous pouvez passer à une commande directe ou télécharger une copie locale.

Vérifiez les fichiers source

Lisez SKILL.md et les fichiers associés affichés par SkillsMP avant de décider de l'installer.

Affichage de SKILL.md

SKILL.md
Instructions source · Aperçu en lecture seule
name
ac-ideate
description
Run Auto-Claude AI-powered ideation for any project directory. Generates low-hanging fruit, high-value features, and UI/UX ideas using the Auto-Claude backend. Use when brainstorming new specs, features, or architectural directions for a project.
# ac-ideate — Auto-Claude Ideation Runner Runs the Auto-Claude ideation pipeline against a project directory and writes spec ideas to `<project>/.auto-claude/ideation/`. ## Binary location ``` /Applications/Auto-Claude.app/Contents/Resources/backend/runners/ideation_runner.py ``` Always set `PYTHONPATH` to the backend root: ``` PYTHONPATH=/Applications/Auto-Claude.app/Contents/Resources/backend ``` ## Basic usage ```bash # Ideate on current directory PYTHONPATH=/Applications/Auto-Claude.app/Contents/Resources/backend \ python3 /Applications/Auto-Claude.app/Contents/Resources/backend/runners/ideation_runner.py \ --project /path/to/project # With options PYTHONPATH=/Applications/Auto-Claude.app/Contents/Resources/backend \ python3 /Applications/Auto-Claude.app/Contents/Resources/backend/runners/ideation_runner.py \ --project /path/to/project \ --types low_hanging_fruit,high_value_features \ --thinking-level high \ --max-ideas 8 \ --refresh ``` ## All flags | Flag | Default | Description | |------|---------|-------------| | `--project PATH` | cwd | Project directory to analyze | | `--output PATH` | `<project>/.auto-claude/ideation` | Where to write results | | `--types LIST` | all | Comma-separated: `code_improvements`, `ui_ux_improvements`, `documentation_gaps`, `security_hardening`, `performance_optimizations`, `code_quality` | | `--thinking-level` | `medium` | `low` / `medium` / `high` | | `--max-ideas N` | 5 | Max ideas per type | | `--model` | `sonnet` | `haiku`, `sonnet`, `opus`, or full model ID | | `--refresh` | false | Force regenerate even if results exist | | `--append` | false | Add to existing session instead of replacing | | `--no-roadmap` | false | Skip roadmap context | | `--no-kanban` | false | Skip kanban context | | `--fast-mode` | false | Faster Opus 4.6 output | ## Open a Zellij tab for ideation ```bash # New tab — runs ideation then drops to shell zellij --session $ZELLIJ_SESSION_NAME action new-tab --name "ideation" \ -- sh -c 'cd /path/to/project && \ PYTHONPATH=/Applications/Auto-Claude.app/Contents/Resources/backend \ python3 /Applications/Auto-Claude.app/Contents/Resources/backend/runners/ideation_runner.py \ --project . --thinking-level high --max-ideas 8 --refresh; exec $SHELL' ``` `$ZELLIJ_SESSION_NAME` is set automatically inside any Zellij session. If running outside, replace with the session name from `zellij list-sessions`. ## Typical workflow 1. **New project** — run with `--types low_hanging_fruit,high_value_features --refresh` 2. **Read results** — `ls <project>/.auto-claude/ideation/` 3. **Promote to specs** — copy ideas into `.auto-claude/specs/<NNN>-<slug>/spec.md` 4. **Dispatch** — `./dispatch.nu run --wave N --yes` or via Auto-Claude UI ## Projects | Project | Path | |---------|------| | dub (EFI bootloader) | `$HOME/efi` | | bop (jobcard system) | `$HOME/bop` | | zam (unikernel VM reader) | `$HOME/zam` | ## Notes - Ideation reads `PLAN.md`, existing specs, and `CLAUDE.md` for context automatically. - `--refresh` is safe to run repeatedly — it overwrites the previous session. - Does **not** require an API key — uses Auto-Claude's OAuth session (Claude Code subscription). - Inside an agent session, `CLAUDECODE` env var is already set; no need to unset it for this runner.
Voir sur GitHub