The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection as it is designed to ingest and analyze external code and error messages which could contain malicious instructions.
Ingestion points: Source code files and error descriptions retrieved via the Read tool and the mcp__julie navigation tools.
Boundary markers: Absent; the workflow does not provide specific delimiters or instructions to the agent to disregard natural language instructions that might be embedded in the code or logs being investigated.
Capability inventory: The skill is granted access to high-privilege tools including Bash, Write, Edit, and Read.
Sanitization: None; the skill processes data from the environment as-is to facilitate debugging.
[COMMAND_EXECUTION]: The workflow explicitly instructs the agent to use the Bash tool to reproduce bugs and run test suites. This is a primary function of the skill and used within the context of a methodical debugging process.
[SAFE]: No obfuscation, hidden URLs, or unauthorized data exfiltration patterns were detected. The use of specialized MCP tools (mcp__sherpa, mcp__julie, mcp__goldfish) represents a standard integration for structured agentic workflows.

bug-detective