The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to the nature of the debugging process.
Ingestion points: The agent is instructed to read and analyze untrusted external content including "captured artifacts (HAR file, log dump, core dump)" in Phase 1 and "error messages" in Phase 2.
Boundary markers: There are no specific instructions or delimiters provided to help the agent distinguish between its instructions and potentially malicious commands embedded within the logs or error messages being analyzed.
Capability inventory: The skill utilizes high-capability tools, including shell execution (bash), network requests (curl, wget), and version control (git bisect), which could be targets for an attacker-controlled log file.
Sanitization: No sanitization or validation of the ingested diagnostic data is performed before the agent processes it.
[COMMAND_EXECUTION]: The skill explicitly directs the agent to execute a wide variety of commands to create feedback loops, including CLI invocations, script execution, and running test harnesses. While this is the intended purpose of the skill, it relies on the agent's ability to safely construct and execute shell commands based on its analysis of the code and bug reports.

diagnose