Security capability SKILL arsenal — 6 independently-loadable security skill modules for AI agents
Extracted and distilled from SecAgent-TUI. Each SKILL is an executable instruction set — an AI agent gains the corresponding capability immediately upon loading.
| SKILL | Capability | Summary |
|---|---|---|
secagent-identity |
Identity | Dual-mode persona + reasoning chain + thinking depth |
secagent-methodology |
Methodology | 4-phase pentest cycle + CVE matching + defense bypass |
secagent-orchestration |
Orchestration | 9 agent types + fan-out + mailbox communication |
secagent-sandbox |
Sandbox | Approval chain + network ACL + execution audit |
secagent-reasoning |
Reasoning | Clue extraction + causal graph + 12 anomaly detectors |
secagent-knowledge |
Knowledge | Local LLM-Wiki + cross-agent sharing + cross-session memory |
# Load a single SKILL in Claude Code
/load_skill secagent-identity
# Or read directly
Skill("secagent-methodology")Any AI agent with file-read capability can load a SKILL.md directly.
- Executable, not documentation — Each SKILL is an instruction set, not an API reference
- Self-bootstrapping — Agent initializes its own environment on first load
- Independent — Each SKILL works standalone, or combined
- Universal — Not tied to any specific language or framework
Extracted from SecAgent-TUI — a deep fork of DeepSeek-TUI built as a multi-agent security research platform.
MIT