Skip to main content

fallback-file-ops

Use run_shell as fallback when execute_code_sandbox or shell_agent fail on filesystem operations

Quellinformationen

Repository
HKUDS/OpenSpace
Letzte Quellaktivität
17. Juli 2026 um 03:43
Erkannte Sprache von SKILL.md
Englisch
Sterne
7.773
Forks
923

Installationsoptionen

Standardmäßig ist der Prompt ausgewählt, der zuerst die Quelle prüft. Sie können zu einem direkten Befehl wechseln oder eine lokale Kopie herunterladen.

Quelldateien prüfen

Lesen Sie SKILL.md und alle von SkillsMP angezeigten Begleitdateien, bevor Sie sich für eine Installation entscheiden.

Datei-Explorer
2 Dateien

SKILL.md wird angezeigt

SKILL.md
Quellanweisungen · Schreibgeschützte Vorschau
name
fallback-file-ops
description
Use run_shell as fallback when execute_code_sandbox or shell_agent fail on filesystem operations
# Fallback File Operations ## Purpose When `execute_code_sandbox` or `shell_agent` fail with 'unknown error' on filesystem operations, fall back to `run_shell` with explicit `mkdir` and file write commands. This hybrid approach is more reliable for creating complex project structures. ## When to Apply - `execute_code_sandbox` returns 'unknown error' on file/directory creation - `shell_agent` fails to write files or create nested directories - You need to create complex project structures with multiple levels ## Procedure ### Step 1: Detect Tool Failure Monitor for these failure patterns: - Error message contains "unknown error" - File operations silently fail (file not created) - Directory creation returns success but directory doesn't exist ### Step 2: Switch to run_shell When failures occur, immediately switch to `run_shell` for filesystem operations: ```bash # Create directory structure explicitly run_shell: mkdir -p /path/to/nested/directory # Write files using shell redirection or echo run_shell: echo "content" > /path/to/file.txt # Or use cat with heredoc for multi-line files run_shell: cat > /path/to/file.txt << 'EOF' line 1 line 2 EOF ``` ### Step 3: Verify Creation After each operation, verify the file/directory was created: ```bash run_shell: ls -la /path/to/created/item run_shell: test -f /path/to/file && echo "File exists" run_shell: test -d /path/to/dir && echo "Directory exists" ``` ### Step 4: Continue with Hybrid Approach - Use `run_shell` for all filesystem operations (mkdir, write, copy, move) - Continue using `execute_code_sandbox` or `shell_agent` for code execution, compilation, or other non-filesystem tasks - Document which tool handles which operation for clarity ## Example: Creating Project Structure ```yaml # Failed attempt with execute_code_sandbox execute_code_sandbox: create directory src/components # Fallback with run_shell run_shell: mkdir -p src/components run_shell: mkdir -p src/utils run_shell: mkdir -p tests/unit run_shell: echo "// Component file" > src/components/Button.tsx run_shell: echo "// Utility file" > src/utils/helpers.ts ``` ## Best Practices 1. **Be explicit**: Always use full paths or confirm working directory 2. **Create parent directories**: Use `mkdir -p` for nested structures 3. **Verify before proceeding**: Check files exist before dependent operations 4. **Log tool switches**: Note when you fall back to run_shell for debugging 5. **Batch operations**: Group related file operations in consecutive run_shell calls ## Common Pitfalls - Don't assume filesystem state after tool failure - Don't mix tools for the same file operation (pick one and stick with it) - Don't skip verification steps when using fallback
Auf GitHub ansehen