The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes shell commands via Bash to manage the file system during migration.
Evidence: Step 4C and 4E use commands like mkdir -p "{project_path}/implementation_process/in_progress/{task_name}" and mv "{project_path}/.../{task_name}.md" ....
Risk: The variables {project_path} and {task_name} are derived from the environment or existing file names. Because these are wrapped in double quotes rather than single quotes in the bash string, shell metacharacters such as $(...) or ` could be evaluated, leading to arbitrary command execution if a file is maliciously named.
[PROMPT_INJECTION]: The skill exhibits an Indirect Prompt Injection surface (Category 8).
Ingestion points: The skill reads untrusted content from existing markdown task files using the Read tool in Step 4A.
Boundary markers: No delimiters or specific instructions (e.g., 'treat the following as data only') are present to isolate the read content from the agent's logic.
Capability inventory: The agent has access to Bash (system commands), Write (file creation), and Read (file access) across all migration scripts.
Sanitization: No validation or escaping is performed on the extracted content before it is interpolated into new file templates in Step 4D.
Risk: Malicious instructions embedded within a task file's content could be interpreted by the agent as instructions during the parsing and migration process, particularly in 'Automatic Mode' where user confirmation is bypassed.

task-folder-migrator