project-directory-migration
Warn
Audited by Gen Agent Trust Hub on Oct 8, 2026
Risk Level: MEDIUMDYNAMIC_EXECUTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DYNAMIC_EXECUTION]: The script scripts/claude-code-migrate.sh generates and executes Python code dynamically using the python3 -c command. It interpolates shell variables directly into Python string literals (e.g., old_dir = '${OLD_DIR}'). Because these variables are not escaped, a directory path containing a single quote can be used to terminate the string and execute arbitrary Python commands. This pattern is present in path resolution, session indexing, and history log processing.
- [COMMAND_EXECUTION]: The skill uses the Bash tool to perform significant system operations, including moving project directories, creating symlinks, and executing environment-specific commands. Notable operations include mv on repository folders and mise trust, which authorizes the execution of configuration files in the new directory. These actions modify the user's filesystem and security state.
- [INDIRECT_PROMPT_INJECTION]: The skill is susceptible to exploitation through malicious directory paths provided as input. 1. Ingestion points: The skill collects source and target paths via the AskUserQuestion tool in SKILL.md. 2. Boundary markers: Absent. No validation or delimiters are used to ensure paths are treated as literal strings. 3. Capability inventory: The skill's migration script has broad file system access, can execute code via Python, and interacts with environment management tools like mise and uv. 4. Sanitization: Absent. The script fails to sanitize or escape input variables before they are used in sensitive shell and Python execution contexts.
Audit Metadata