The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of shell commands, specifically grep for pattern matching within local files and curl for making network requests to test authentication endpoints. These tools are used as intended for security auditing and traffic analysis.
[INDIRECT_PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection as it processes untrusted network traffic captured in log.txt. Maliciously crafted HTTP traffic stored in these logs could potentially contain instructions aimed at influencing the agent's behavior or reporting during the analysis phase.
Ingestion points: Reads network traffic data from the local log.txt file using various grep commands (e.g., searching for JWTs, session tokens, and roles).
Boundary markers: No explicit boundary markers or instructions to ignore embedded content are defined for the data being parsed.
Capability inventory: The skill has access to shell command execution (grep, curl), file system reading, and network communication.
Sanitization: The skill does not perform sanitization or validation of the data extracted from the logs before it is processed by the agent.

mitm-find-auth