The Agent Skills Directory

[PROMPT_INJECTION]: The skill contains an instruction to suppress reporting and conclusions after the task is finished ("После завершения всех изменений не предоставляй отчёт и не делай выводов"). This is a concealment tactic that reduces user oversight of the agent's actions.
[COMMAND_EXECUTION]: The instructions explicitly direct the agent to prefer shell commands, specifically heredoc via cat, for all file operations ("При записи/перезаписи файлов предпочитай shell-команды (heredoc через cat)") instead of built-in editor tools.
[PROMPT_INJECTION]: The skill uses broad, permissive language that could be interpreted as bypassing implicit constraints ("Разрешаю использовать любые инструменты и действия для выполнения задачи").
[PROMPT_INJECTION]: The skill is vulnerable to Indirect Prompt Injection (Category 8):
Ingestion points: The skill reads and processes all *.php files within specific component directories.
Boundary markers: No boundary markers or instructions to ignore embedded instructions in the source files are provided.
Capability inventory: The skill has the capability to execute shell commands (cat, pwd), read files, and write/modify files across the project directory.
Sanitization: There is no evidence of sanitization or validation of the content being read from the PHP files before it is processed or used in shell commands.

bx-trf