The Agent Skills Directory

[Prompt Injection] (SAFE): No instructions designed to override AI safety filters or hijack agent behavior were found. The instructional logic is focused on classification tasks.
[Data Exposure & Exfiltration] (SAFE): The skill does not access sensitive local credentials (e.g., ~/.ssh) or attempt to transmit data to external domains. It strictly analyzes repository file paths to determine change types.
[Remote Code Execution] (SAFE): There are no patterns involving remote script downloads or piped execution from the internet.
[Obfuscation] (SAFE): The file contains clear, human-readable markdown and YAML frontmatter with no encoded strings or hidden characters.
[Indirect Prompt Injection] (LOW): This skill has an inherent ingestion surface as it processes untrusted file paths and code diffs.
Ingestion points: The skill reads file names and diff content from the current PR or work set.
Boundary markers: No explicit delimiters or instructions to ignore embedded content are used.
Capability inventory: The skill is limited to providing text-based recommendations and does not have built-in file-write or network-access capabilities.
Sanitization: No specific sanitization of the input data is performed. Given the advisory nature of the output, the risk is minimal.

change-type