Skip to main content
prompt GP ChatGPT

storage_access — Claude Code prompt: file access rules

storage_access is a Claude Code prompt that limits an agent’s access to stored resources. It defines allowed paths for robot references, observations, and session work.

MIT ★ 488 updated 2026-11-15

Install into Claude Code · Manually: put the files in place All methods and platforms ↓
.claude/commands/storage_access.md
Open on GitHub

What storage_access does

storage_access sets which files an agent may read and where it can save its own session materials. Skill documentation is read through declared tools; updates are allowed only through the declared writing tool when enabled.

Images in {{ images_dir }} are read-only. The same restriction applies to exact paths supplied in the initial input or tool results: parent folders, simulator internals, and other task assets are outside the permitted scope. Agents can use {{ scratch_dir }} for session scripts, notes, and calculations. Task files and directories beyond these boundaries cannot be read, searched, or changed.

What it helps with

  • limit reads of observation images
  • restrict robot resource access to supplied paths
  • block searches through other task directories
  • keep session scripts and notes in a designated folder

Example prompts for the agent

  • Use storage_access to read the observation at the supplied path.
  • Apply storage_access and save the session calculations in {{ scratch_dir }}.
  • Follow storage_access and avoid searching for task files outside the permitted paths.

How to install storage_access

Manually: put the files in place

.claude/commands/storage_access.md

Questions about storage_access

How do I install storage_access in Claude Code?

Save the prompt as .claude/commands/storage_access.md.

What does storage_access do?

It defines permitted read and write locations for robot resources, observation images, and session files. Other task files remain out of scope.

Which files can storage_access read?

It allows reads from {{ images_dir }} and from exact observation or robot resource paths supplied in the initial input or tool results. Skill documentation and robot docs are accessed through declared tools.

Source text

storage_access.md · MIT
If you need to access stored resources, limit access to the following:

- Skills and robot documentation through the declared skill listing and reading
  tools; skill updates only through the declared skill-writing tool when enabled.
- Observation images in `{{ images_dir }}` (read-only).
- Exact files supplied in initial input or tool results as `observation_path`,
  `observation.artifact_ref`, camera `image_path`, `robot_calibration_path`, or
  entries in `robot_resources` (read-only). These are
  published observations and robot references, not permission to inspect their
  parent directories, simulator internals, or task assets.
- `{{ scratch_dir }}` for reading and writing your own scripts, notes, and
  calculation results for this session.

Do not read, search, or modify task-related files or directories outside these
scopes.

© Open-RUA · MIT · Open-RUA/uni-robot-agent

Similar prompts

Prompts catalog →

From the same set41

The whole set →
resource_global-serial The resource_global-serial Claude Code prompt sequences tool calls that share one exclusive resource. It batches calls whose outcomes are already determined and separates steps that need feedback. GP ChatGPT ★ 488repo stars resource_per-resource The resource_per-resource prompt for Claude Code and ChatGPT assigns tool calls to declared resource domains. It orders physical-world calls while allowing independent snapshot perception to overlap. GP ChatGPT ★ 488repo stars return_immediate The return_immediate Claude Code prompt sets how an agent handles tool calls: each result arrives before planning continues. It groups independent calls and pauses when the next decision needs fresh feedback. GP ChatGPT ★ 488repo stars robodojo_main The robodojo_main prompt for Claude Code helps verify RoboDojo actions against environment snapshots. It separates intermediate changes from evidence of a completed action. GP ChatGPT ★ 488repo stars tool_calling The tool_calling prompt for Claude Code and ChatGPT sets rules for using structured tools. It guides the agent to select a declared tool by its description and parameter schema. GP ChatGPT ★ 488repo stars batch The batch prompt for Claude Code groups independent reads and perception queries in one model response. Physical actions stay ordered, with a pause where their result needs checking. GP ChatGPT ★ 488repo stars hybrid The hybrid Claude Code prompt coordinates a humanoid as one actor. It combines upper-body targets with walking velocity while native control handles the legs and waist. GP ChatGPT ★ 488repo stars hybrid The hybrid prompt for Claude Code reports RoboDojo task progress with evidence-backed subgoals. A reviewer checks action-model proposals before execution. GP ChatGPT ★ 488repo stars

Project sponsors

Companies that keep this analytics open