CtrlK
BlogDocsLog inGet started
Tessl Logo

filesystem_guide

Guidance for workspace file tools vs shell commands.

Invalid
This skill can't be scored yet
Validation errors are blocking scoring. Review and fix them to unlock Quality, Impact and Security scores. See what needs fixing →
SKILL.md
Quality
Evals
Security
  • FILESYSTEM vs SHELL ROUTING (strict rules):
    • The workspace_* file tools (workspace_read_file, workspace_list_directory, workspace_write_file, etc.) ONLY work inside the configured workspace folder. They CANNOT access any other location on the computer. workspace_read_file also supports image files — it displays them inline in chat. To also analyze image contents, use analyze_image with source='file'.
    • For ANY path outside the workspace (D:, C:\Users, /home, etc.), ANY terminal command (python --version, git status, pip install, etc.), or ANY system operation: you MUST use run_command. No exceptions.
    • NEVER tell the user to open a terminal or run a command themselves. You have run_command — use it directly.
  • ORDINARY SAVES AND CONVERSATIONAL CONFIRMATION:
    • Creating, writing, copying, and exporting ordinary workspace files are not approval-gated operations. Use the relevant workspace tool directly when the user asks to save without requesting a separate confirmation.
    • If the user explicitly says to ask before saving, treat that as normal conversation, not a protected-action approval: finish and present the material, ask for confirmation, and make zero write/copy/export calls on that turn. After a later affirmative user message, make exactly one normal write/copy/export call and report success only after the tool succeeds.
    • Do not create an approval request or promise an approval card for an ordinary workspace save. Move and delete operations keep their existing configured approval behavior.
Repository
siddsachar/row-bot
Path

tool_guides/filesystem_guide/SKILL.md

Commit

b4dcda4

Last updated
First committed

Is this your skill?

If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.

Claim skill