robsonrung/rar-skills

dcode-runner

Execute prompts using DeepAgents CLI (dcode) non-interactive mode with the user's already-configured model and credentials.

Quelltext ansehen

Verwendung

  1. Installationsbefehl kopieren und das Quellpaket vor der Installation in der eigenen Agentenumgebung prüfen.
  2. Benötigte Werkzeuge, Zugangsdaten und Hosts im Original prüfen.
  3. Eingabe und gewünschtes Ergebnis beschreiben und die Ausgabe mit den Quellen abgleichen.
npx skills add https://github.com/robsonrung/rar-skills --skill dcode-runner

Vor dem Start

  • Die Installation umfasst weder Modellabo noch API-Kontingent oder Dateizugriffsrechte.
  • Sterne und Installationen zeigen Aktivität, keine unabhängige Qualitäts- oder Sicherheitsprüfung.
Originales Skill-Dokument

Aus dem Quell-Repository gerendert; Überschriften, Beispiele, Code, Tabellen, Links und Bilder bleiben erhalten.

Dcode Runner

Execute the caller's scoped prompt through the local CLI. Shared roles, envelope keys, and result handling live in shared/references/runner-common.md. Preserve seat fidelity: name the provider that actually answered, and report unavailable or unverified seats honestly.

Invocation

Set SKILL_DIR to this loaded skill directory in the same shell call:

bash
SKILL_DIR="<absolute path of this skill directory>";
python3 "$SKILL_DIR/scripts/run_dcode.py" "<scoped prompt>" --json

An approved workflow supplies the exact model, effort, role, tool mode, and receipt policy. Pass those approved options and --disable-fallback; this basic example does not authorize a different route.

Runtime boundaries

  • Use only for an explicitly requested manual dcode run. This wrapper cannot bind an exact model and cannot be an approved implementation or review route.
  • The user's existing model and credentials stay configured in dcode. --model is a metadata label and is not forwarded.
  • Permission checks stay enabled unless an authorized call passes --auto-approve. Prompt and file context may be sent to the configured provider.
  • Analysis roles use a read-only prompt overlay, not an enforced sandbox. Isolate the workspace for untrusted input; do not treat --restrict-tools as a security boundary.
  • Direct calls have a labelled fallback chain when the CLI is missing. Do not attribute a fallback answer to dcode. Use --disable-fallback when the caller requires that seat.
  • Native output has no session ID. Resuming the latest session uses shared state and must not run concurrently against the same home.

Load by need

Result

Read agent_message and the wrapper envelope. Check success, serving-model receipt, effective runner/provider, and any fallback or malformed-output status. A completed process alone does not prove the requested task or model was used. Report touched files when the caller authorized edits.

Integration

agents/openai.yaml exposes this skill as a native Codex-app subagent seat; do not remove it.

Verwandte Aufgaben

aus demselben Repository

Weitere Skills

Alle Skills