The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes a user-provided 'brief' to generate social media copy. While this represents a standard data ingestion surface, the skill's capabilities are limited to text generation and reading local reference files. There are no high-risk capabilities such as file system writes, network requests, or command execution that could be exploited via malicious input in the brief.
[REMOTE_CODE_EXECUTION]: No remote code execution patterns were detected. The skill relies entirely on local markdown reference files provided within the skill's directory structure.
[DATA_EXFILTRATION]: There are no network operations (e.g., curl, fetch, or requests) or attempts to access sensitive system files (e.g., .ssh, .env). The skill operates within a contained environment focusing on text transformation.
[PROMPT_INJECTION]: The instructions in SKILL.md use standard, non-adversarial instructional language. It includes a 'De-slop pass' designed to improve the quality and human-like nature of the output, which is a benign quality-control measure rather than a bypass of safety guidelines.

social-copy