Detect prompt injection and data leakage risks in AI-readable skill files.
npx skills add https://github.com/picahq/skills --skill skill-sanitizer
Please help me install this Agent Skill. Skill: skill-sanitizer Source: https://github.com/picahq/skills/tree/main/skills/skill-sanitizer Command: npx skills add https://github.com/picahq/skills --skill skill-sanitizer
This Skill provides a security review process to detect prompt injection, hidden instructions, and other attack patterns before a skill is trusted.
Run this Skill against a candidate skill file to generate a sanitization report. Review the findings and apply remediation before deployment.
High-intent search queries and answers about installing and using this skill.
Detect prompt injection in AI-readable skill files by applying static analysis to identify malicious markers, hidden instructions, and obfuscated content. This process generates a formal sanitization report to review before deployment.
Static analysis for AI security is the process of scrutinizing skill files, code blocks, and hidden content like HTML comments to find attack patterns. You need it before deploying any AI skill to prevent data leakage and injection.
Check AI skill files for hidden content and data leakage risks by validating URLs, inspecting data flows, and scanning for zero-width characters or Markdown comments. This scrutiny prevents unauthorized exfiltration during skill execution.
Yes, you can sanitize AI skills using predefined patterns and allowlists to systematically validate content. This step-by-step procedure identifies security vulnerabilities and triggers a formal review before you trust the skill.
Static analysis for prompt injection detection is limited to predefined patterns, URL validation, and hidden-content checks within skill files. It may not catch novel or dynamic attack vectors that emerge outside these established static rules.
mitul-bhatiaHarden Perl applications with taint mode, input validation, and parameterized DBI queries.
Snowflake-LabsA concise summary of your roleplay character: name,</name> personality, and/or behavior, formatted as a single string sentence. Typically不超过50字，此处省略
mvillmowDetect security vulnerabilities and hardcoded secrets in code.
Zhow01Detect and block unsafe prompts and secret leakage in OpenClaw deployments.
