The Agent Skills Directory

[SAFE]: No malicious patterns or security vulnerabilities were detected. The skill is designed for benign text manipulation within the agent's environment.
[PROMPT_INJECTION]: No instructions to bypass safety filters or override system behavior were found. The skill processes untrusted input only for rewriting purposes as requested by the user.
[DATA_EXFILTRATION]: The skill does not access sensitive files, environment variables, or perform network operations.
[SAFE]: Analysis of indirect prompt injection surface:
Ingestion points: User input provided via the /sharpen command or preceding messages.
Boundary markers: No specific delimiters are used for the output prompt, though instructions specify a clean block.
Capability inventory: Limited to text output and the standard AskUserQuestion tool; no access to shell, filesystem, or network.
Sanitization: No input sanitization is performed, but the risk is mitigated by the lack of dangerous capabilities available to the skill.

sharpen