The Agent Skills Directory

Remote Code Execution (Untrusted Source): The skill's primary function is to extract and execute code snippets provided by external issue reporters. Executing code from unverified third parties is a notable security consideration. This risk is addressed by a mandatory human confirmation step (Step 5.5) before any execution occurs, and a requirement to verify environment isolation (Step 0).
Indirect Prompt Injection Surface: The skill ingests data from external issue trackers (descriptions, comments, and attachments), which could contain malicious instructions designed to manipulate the agent. The skill explicitly directs the agent to treat this content strictly as data and provides clear warnings regarding potential injection attempts.
Command Execution: The skill performs shell command execution to build and run project-specific reproducers using tools like Gradle, Pip, or Make. While these operations are necessary for the skill's purpose, they represent a potential attack surface if the environment is not properly secured.
Obfuscation (Terminal Escape Sequences): The skill uses OSC 8 terminal escape sequences to create clickable links in terminal output. Although these are non-standard sequences, they are used for functional improvements in terminal readability rather than for concealing malicious intent.

reproducer