The Agent Skills Directory

[SAFE] (SAFE): The skill contains no executable scripts or malicious commands. It consists of instructional markdown and reference data for code debugging.
[Indirect Prompt Injection] (SAFE): The skill is designed to process untrusted external data (error messages and code in variables $0 and $1) and instructs the agent to execute modified versions of that code. This creates a vulnerability surface for indirect prompt injection where an attacker could influence the agent's code execution through malformed error reports or source code. However, this risk is inherent to the skill's primary purpose. Evidence Chain: 1. Ingestion points: $0 and $1 in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: Fix-and-rerun logic implies subprocess execution or code evaluation. 4. Sanitization: Absent.
[Prompt Injection] (SAFE): The reference file contains role-play instructions ('You are a code repair specialist'), but these are limited to the intended functional scope of the skill and do not attempt to bypass safety filters or extract system prompts.

code-debugging