The Agent Skills Directory

[SAFE]: No malicious patterns, obfuscation, or deceptive behaviors were detected in the skill files. The instructions are transparent and aligned with the stated purpose of prompt engineering.
[DATA_EXPOSURE]: No hardcoded credentials, API keys, or attempts to access sensitive system files (e.g., SSH keys, environment variables) were identified.
[REMOTE_CODE_EXECUTION]: The skill does not perform runtime remote code execution or download external scripts. Installation via Git clone from the author's repository is a standard, user-initiated setup procedure.
[INDIRECT_PROMPT_INJECTION]: The skill handles untrusted data via {{USER_REQUEST}} and {{INPUT_CONTEXT}}. However, it includes a dedicated 'Input sanitization' constraint that instructs the agent to check for injection patterns and flag them before processing, mitigating the risk of indirect prompt injection.
[COMMAND_EXECUTION]: There are no high-risk shell commands or privilege escalation attempts. The optimization 'hacks' described (e.g., disabling unused tools, limiting sub-agents) are security-positive configurations that promote the principle of least privilege.

promptor