The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to write and execute shell scripts (e.g., /tmp/debug_api.sh) using the Bash tool and chmod +x. This allows the agent to perform arbitrary system operations during the diagnostic process.
[REMOTE_CODE_EXECUTION]: The agent is directed to generate and run Python scripts (e.g., /tmp/analyze_db.py) using poetry run python. This involves dynamic code generation where the logic is created at runtime based on the agent's interpretation of a problem.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8). It ingests untrusted data in the form of user-provided bug reports and uses that input to influence the creation of executable diagnostic scripts.
Ingestion points: User-provided error descriptions and bug reports (triggered by keywords like 'bug' or 'error').
Boundary markers: None identified. The skill lacks instructions to sanitize user input or ignore instructions embedded within the bug reports.
Capability inventory: The agent has access to Bash, Write, Read, and Edit tools, allowing it to create, modify, and execute files across the system.
Sanitization: No evidence of input validation or escaping of user-provided content before it is processed by the agent's reasoning loop.
[DATA_EXPOSURE]: The skill contains hardcoded absolute file paths (e.g., /Users/young/project/career_ios_backend). This discloses the host machine's username and internal directory structure, which can be used for further reconnaissance.

debugging