Skip to main content

env-abstraction

Handles abstraction of environment-specific commands between Jetski and Gemini-CLI. Use when switching execution context between local terminal environments. Do not use for debugging logic issues.

Informations de source

Dépôt
nodejs/node
Dernière activité de la source
4 octobre 2026 à 21:04
Langue détectée de SKILL.md
anglais
Étoiles
122 301
Forks
39 450

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
env-abstraction
description
Handles abstraction of environment-specific commands between Jetski and Gemini-CLI. Use when switching execution context between local terminal environments. Do not use for debugging logic issues.
# Environment Abstraction Skill This skill provides a layer of abstraction over environment-specific tools and commands, allowing skills to be reused across different platforms like `jetski` and `gemini-cli`. ## Core Principles 1. **Decoupling**: Keep tool calls independent from platform specific implementations inside general workflow skills. 2. **Detection**: Determine the environment at startup (e.g., via environment variables or tool availability). 3. **Mapping**: Map abstract actions to concrete commands or tool calls based on the environment. ## Environment Definitions ### Jetski Environment - **Characteristics**: Rich set of specialized tools (e.g., `code_search`, `moma_search`, `gdb-mcp`, `v8-utils`). - **Usage**: Prefer specialized tools over generic shell commands to maximize efficiency and respect workspace constraints (e.g., avoid heavy searches in large workspaces). ### Gemini-CLI Environment - **Characteristics**: May rely more on standard shell commands or a different set of MCP servers. - **Usage**: Fall back to standard tools like `grep`, `find`, or standard `gdb` via `run_command` if specialized Jetski tools are unavailable. ## Abstract Action Mapping When performing common actions, refer to this mapping to choose the correct tool: | Abstract Action | Jetski Implementation | Gemini-CLI Implementation (Fallback) | | :--- | :--- | :--- | | **Code Search** | `code_search` | `grep_search` (if outside Google3) or standard `grep` | | **File Search** | `find_by_name` (if safe) | Standard `find` or `fd` | | **Debugging** | `gdb-mcp` | Standard `gdb` via `run_command` | | **Building** | `tools/dev/gm.py` (use `use_remoteexec=true`) | `tools/dev/gm.py` | | **Testing** | `tools/run-tests.py` | `tools/run-tests.py` | | **Starting Agents** | `invoke_subagent` (tool) | `agentapi new-conversation` (CLI) | ## Implementation Guidelines - When writing or updating skills, refer to actions by their abstract names. - Check for the existence of specialized tools before falling back to generic commands. - Document any environment-specific assumptions in the skill file.
Voir sur GitHub