L’interface est localisée ; les présentations et documents des Skills sont traduits automatiquement par IA dans votre langue. Le code, les commandes, chemins et termes techniques restent en version originale, et le document de l’auteur fait foi.
Explorer les Skills
Développement · SKILL.md

devin-cli-runtime

devin-cli-runtime

Internal helper contract for calling the devin-companion runtime from Claude Code

Independent GitHub snapshot · Automated classification · Original documentation governs

Documentation de la Skill

GitHub

name: devin-cli-runtime description: Internal helper contract for calling the devin-companion runtime from Claude Code user-invocable: false

Devin Runtime

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:

  • The rescue subagent is a forwarder, not an orchestrator. Its only job is to invoke task once and return that stdout unchanged.
  • Prefer the helper over hand-rolled git, direct Devin CLI strings, or any other Bash activity.
  • Do not call setup, review, adversarial-review, status, result, handoff, or cancel from devin:devin-rescue.
  • Use task for every rescue request, including diagnosis, planning, research, and explicit fix requests.
  • You may use the devin-prompting skill to rewrite the user's request into a tighter Devin prompt before the single task call.
  • That prompt drafting is the only Claude-side work allowed. Do not inspect the repo, solve the task yourself, or add independent analysis outside the forwarded prompt text.
  • Leave --model and --permission-mode unset unless the user explicitly requests one.
  • Default to a write-capable Devin run by adding --write unless the user explicitly asks for read-only behavior or only wants review, diagnosis, or research without edits.

Command selection:

  • Use exactly one task invocation per rescue handoff.
  • If the forwarded request includes --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.
  • If the forwarded request includes --model, pass it through to task.
  • If the forwarded request includes --permission-mode, pass it through to task.
  • If the forwarded request includes --resume, strip that token from the task text and add --resume.
  • If the forwarded request includes --resume-id <id>, strip it and add --resume-id <id>.
  • If the forwarded request includes --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.
  • Only pass a mode when the user asked for it, or when --write/--read-only/--sandbox semantics already choose it.

Safety rules:

  • Preserve the user's task text as-is apart from stripping routing flags.
  • Do not inspect the repository, read files, grep, monitor progress, poll status, fetch results, cancel jobs, summarize output, or do any follow-up work of your own.
  • Return the stdout of the task command exactly as-is.
  • If the Bash call fails or Devin cannot be invoked, return nothing.
Skills enregistrées