The Agent Skills Directory

[SAFE]: The skill does not contain any executable scripts or code blocks. It is entirely comprised of natural language instructions and logic tables.
[SAFE]: No hardcoded credentials, API keys, or sensitive file paths (such as .ssh or .env) were detected.
[SAFE]: The skill does not perform any network operations (curl, wget, fetch) or external data transfers.
[SAFE]: No obfuscation techniques (Base64, hex encoding, or hidden Unicode characters) were found.
[PROMPT_INJECTION]: The skill contains defensive instructions against manipulation. The 'Common Bypasses' section explicitly instructs the agent to ignore user attempts to override the 'stop' boundary or force automatic skill execution. These are safe, protocol-enforcing instructions.
[INDIRECT_PROMPT_INJECTION]: While the skill processes information about development issues (which could be sourced from untrusted PR bodies or code), it has no exploitable capabilities.
Ingestion points: Analyzes descriptions of code, specs, and requirements provided in the agent's context (SKILL.md).
Boundary markers: Strong delimiters are used, and the skill enforces a mandatory 'stop' action after outputting its routing decision.
Capability inventory: No tools are used, and the skill explicitly forbids file modifications, network calls, or automatic skill invocation.
Sanitization: Not applicable as the output is constrained to a fixed Markdown template with no dynamic code generation.

route-problem