The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses node to execute local scripts (doc-lookup.cjs and context7-api.cjs) with arguments such as <library> and <question> derived from external context. This command-line interpolation is a potential vector for command injection; if these input variables contain shell metacharacters like semicolons, backticks, or pipes, they could be used to execute arbitrary code on the underlying system.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because its core functionality involves following instructions provided in an external "Bug list" and modifying source code based on "suggested fix directions."
Ingestion points: The bug list (specifically BUG-IDs, descriptions, and fix directions) and the source files within the assigned directory.
Boundary markers: The skill does not define clear delimiters or instructions to ignore malicious commands that might be embedded within the bug descriptions or the code being repaired.
Capability inventory: The skill possesses the capability to modify arbitrary files in the project and execute shell commands via Node.js.
Sanitization: There is no evidence of sanitization or verification logic to ensure that the "suggested fix direction" provided by the potentially untrusted Referee agent does not contain malicious instructions meant to subvert the agent's behavior.

fixer