The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill includes documentation and Indicator of Compromise (IoC) patterns for detecting remote code execution, such as python -c "$(curl ...)", curl | bash, and eval $(curl). These serve as reference data for the agent to identify malicious behavior in audited skills.
[COMMAND_EXECUTION]: The tool utilizes powerful system utilities for its auditing and malware eradication recipes, including find, grep, perl, systemctl, and launchctl. These commands are used to inspect the local filesystem, verify file integrity, and manage system service states.
[DATA_EXFILTRATION]: The skill is designed to identify and scan for sensitive files commonly targeted in supply-chain attacks, including SSH keys (~/.ssh), cloud credentials (~/.aws/credentials), and registry tokens (~/.npmrc). This functionality is limited to its security auditing and recovery workflows.
[OBFUSCATION]: The file reference/unicode-tag-scan.md contains examples of hidden characters, including Unicode Tags (U+E0000–U+E007F), bidi overrides, and zero-width characters. These are provided as specific test cases and reference examples for the skill's standalone Unicode scanner.
[INDIRECT_PROMPT_INJECTION]: This skill possesses a significant attack surface as it ingests untrusted third-party data from skills and marketplace plugins. It mitigates this risk by enforcing strict codepoint scanning, requiring manifest generation with SHA-256 hashes, and using explicit boundary markers for processed content.
[PRIVILEGE_ESCALATION]: The skill provides procedures for managing system persistence mechanisms, such as macOS LaunchAgents and Linux systemd user units. These operations are part of its eradication playbook for removing confirmed malware from the host environment.

chain