brainstorming
brainstorming
You MUST use this before any creative work - creating features, building components, adding functionality, modifying behavior, or planning anything new, in software or out of it (a talk, a business, a renovation).
devin-cli-runtimeInternal helper contract for calling the devin-companion runtime from Claude Code
Independent GitHub snapshot · Automated classification · Original documentation governs
Use this skill only inside the devin:devin-rescue subagent.
Primary helper:
node "${CLAUDE_PLUGIN_ROOT}/scripts/devin-companion.mjs" task "<raw arguments>"Execution rules:
task once and return that stdout unchanged.git, direct Devin CLI strings, or any other Bash activity.setup, review, adversarial-review, status, result, handoff, or cancel from devin:devin-rescue.task for every rescue request, including diagnosis, planning, research, and explicit fix requests.devin-prompting skill to rewrite the user's request into a tighter Devin prompt before the single task call.--model and --permission-mode unset unless the user explicitly requests one.--write unless the user explicitly asks for read-only behavior or only wants review, diagnosis, or research without edits.Command selection:
task invocation per rescue handoff.--background or --wait, treat that as Claude-side execution control only. Strip it before calling task, and do not treat it as part of the natural-language task text.--model, pass it through to task.--permission-mode, pass it through to task.--resume, strip that token from the task text and add --resume.--resume-id <id>, strip it and add --resume-id <id>.--fresh, strip that token and do not add a resume flag.--resume: continue the most recent Devin session in this directory (devin -p -c).--resume-id <id>: resume a specific Devin session (devin -p -r <id>).--fresh: always use a fresh task run, even if the request sounds like a follow-up.task --resume: internal helper for "keep going", "resume", "apply the top fix", or "dig deeper" after a previous rescue run.Permission modes (--permission-mode):
normal: read-only-ish in -p mode — tool calls needing confirmation are auto-rejected. Use for research/diagnosis.accept-edits: file edits auto-approved; shell commands still rejected. Default for --write.smart: a fast model judges each action; rolling out gradually, may not exist on all accounts.dangerous (aliases yolo, bypass): auto-approves everything including shell commands.autonomous: only with --sandbox; commands run inside an OS-level sandbox.--write/--read-only/--sandbox semantics already choose it.Safety rules:
task command exactly as-is.