The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/debug-session-logger.sh is vulnerable to path traversal via the --session (or -s) command-line argument. The script concatenates the user-provided filename with a local directory path without validation, allowing an attacker to read or overwrite sensitive files outside the intended directory (e.g., -s ../../.ssh/authorized_keys).
[COMMAND_EXECUTION]: The initialize_session function in scripts/debug-session-logger.sh is susceptible to shell command injection. It utilizes an unquoted heredoc (<< EOF) to process the $ISSUE_TITLE variable, which causes the shell to evaluate and execute any embedded command substitutions, such as backticks or $() syntax, present in the title string.
[DATA_EXFILTRATION]: The path traversal vulnerability in scripts/debug-session-logger.sh allows for unauthorized data exposure. The resume_session function displays the entire contents of the file specified by the --session argument to standard output, which could be used to leak sensitive information like private keys, configuration files, or environment secrets if an attacker-controlled path is provided.
[EXTERNAL_DOWNLOADS]: The SKILL.md documentation references several third-party Python libraries, including structlog, line_profiler, and requests. While these are standard development tools, they represent external dependencies that the agent may be prompted to install.

Systematic Debugging