The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill contains a specific instruction in SKILL.md to run env | grep -i "DB_\|API_\|SECRET_" | sort. This command is designed to extract and display sensitive credentials, including API keys and database secrets, directly into the agent's output context.
[DATA_EXFILTRATION]: The practice of dumping environment variables containing secrets into the terminal output constitutes a significant data exposure risk. This sensitive information could be captured in logs, shared accidentally, or intercepted by malicious actors monitoring the session.
[COMMAND_EXECUTION]: The skill promotes the execution of powerful shell commands, including psql for direct database interaction and env for system inspection. While intended for debugging, these provide a broad capability for an agent to interact with and expose the underlying host's configuration.
[PROMPT_INJECTION]: The skill exhibits a vulnerability to indirect prompt injection by design.
Ingestion points: The agent is instructed to analyze logs/app.log using grep to find errors and stack traces.
Boundary markers: There are no delimiters or specific instructions (e.g., "treat the following content as data only") used when reading external log files.
Capability inventory: The skill allows access to Bash, psql, curl, and Git tools, which an attacker could attempt to manipulate by injecting malicious instructions into the application logs.
Sanitization: No sanitization or filtering of the log content is performed before the agent processes it.

service-debugging