The Agent Skills Directory

[DATA_EXFILTRATION]: Provides lists of highly sensitive system file paths as targets for the verification of directory traversal vulnerabilities during security assessments.
Evidence: Includes paths such as /etc/shadow, /etc/passwd, /home/$USER/.ssh/id_rsa, /etc/mysql/my.cnf, and Kubernetes service account tokens (/run/secrets/kubernetes.io/serviceaccount/token).
[EXTERNAL_DOWNLOADS]: Skill instructions reference third-party security research tools and payload libraries hosted on external code repositories for use in exploitation workflows.
Evidence: Links to repositories for dotdotpwn, IIS-ShortName-Scanner, shortscan, and windowsblindread for tool acquisition and reference.
[COMMAND_EXECUTION]: Provides specific command-line instructions for executing third-party security scanners and perl scripts.
Evidence: Sample commands provided for perl dotdotpwn.pl, java -jar ./iis_shortname_scanner.jar, and shortscan.
[OBFUSCATION]: Documents numerous encoding and filter bypass techniques used to conceal path traversal characters from automated security filters.
Evidence: Enumerates URL encoding (%2e), Double URL encoding (%252e), Unicode encoding (%u002e), and overlong UTF-8 sequences (%c0%ae).
[INDIRECT_PROMPT_INJECTION]: The skill's workflow involves processing data from external network targets which could contain adversarial content designed to influence the agent's behavior during automated scanning.
Ingestion points: Network responses captured via curl, HTTP logs, and scanner output referenced in the workflow (SKILL.md).
Boundary markers: Absent.
Capability inventory: Network requests (curl), subprocess execution (perl, java), and target file verification.
Sanitization: Absent.

payloads-path-traversal