The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The file reference/auth-deser.md contains several hardcoded default credentials used for testing, including 'admin:admin', 'admin:password', and 'root:root'.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: SKILL.md instructs the agent to use curl, gobuster, ffuf, and whatweb to retrieve data from external web targets, which is then processed in the agent context.
Boundary markers: No boundary markers or instructions to ignore embedded commands in the target data are provided.
Capability inventory: The skill is granted access to powerful tools including Bash, Write, and Read in the YAML frontmatter.
Sanitization: No sanitization or validation of the fetched external content is described before the data is interpreted by the agent.
[COMMAND_EXECUTION]: The skill workflow and references encourage the execution of complex shell commands and the use of automated exploitation tools like sqlmap, commix, and tplmap which interact with untrusted inputs.
[OBFUSCATION]: Several reference files (reference/xss.md, reference/auth-deser.md, reference/ssrf-lfi.md) include encoded content such as Base64-encoded serialization payloads, HTML entities (e.g., a), and hex-encoded IP addresses (0x7f000001) intended to bypass security filters.
[DATA_EXFILTRATION]: Documentation in reference/xss.md provides explicit examples of how to exfiltrate sensitive data, such as document.cookie, to external domains via script injections.
[EXTERNAL_DOWNLOADS]: The skill references and relies on multiple third-party tools and resources, including gobuster, ffuf, whatweb, sqlmap, commix, tplmap, jwt_tool, phpggc, and ysoserial.jar.

web-security