The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous templates and generation logic for reverse and bind shells designed to establish remote control of target systems.
Evidence: SKILL.md includes interactive shell commands for Bash, Python, and PowerShell.
Evidence: scripts/payload_generator.py contains logic to generate one-line shells for languages like Perl, Ruby, and PHP.
[DYNAMIC_EXECUTION]: The skill involves the creation of executable shellcode and exploit payloads.
Evidence: Instructions utilize pwntools for assembly (asm) and buffer overflow exploitation.
Evidence: scripts/payload_generator.py programmatically constructs web shells and injection payloads.
[EXTERNAL_DOWNLOADS]: The skill requires the installation of external Python libraries through standard registries.
Evidence: Instructions specify pip install pwntools keystone-engine capstone.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data such as CVE advisories and vulnerability research, which could contain malicious instructions.
Ingestion points: Researching CVEs and official advisories (SKILL.md).
Boundary markers: The skill encourages structured reporting but does not define strict delimiters for external content.
Capability inventory: High-privilege actions including network operations (reverse shells), file system manipulation (web shells), and command execution.
Sanitization: None explicitly defined for external advisory content.

03-exploit-development