JesusConwellpy/secagent-skills

# SecAgent Skills Extracted and distilled from [SecAgent-TUI](https://github.com/JesusConwellpy/SecAgent-TUI). Each SKILL is an **executable instruction set** — an AI agent gains the corresponding capability immediately upon loading. [中文说明](./README.zh-CN.md) ## SKILL Index | SKILL | Capability | Summary | |-------|-----------|---------| | [`secagent-identity`](./secagent-identity/SKILL.md) | Identity | Dual-mode persona + reasoning chain + thinking depth | | [`secagent-methodology`](./secagent-methodology/SKILL.md) | Methodology | 4-phase pentest cycle + CVE matching + defense bypass | | [`secagent-orchestration`](./secagent-orchestration/SKILL.md) | Orchestration | 9 agent types + fan-out + mailbox communication | | [`secagent-sandbox`](./secagent-sandbox/SKILL.md) | Sandbox | Approval chain + network ACL + execution audit | | [`secagent-reasoning`](./secagent-reasoning/SKILL.md) | Reasoning | Clue extraction + causal graph + 12 anomaly detectors | | [`secagent-knowledge`](./secagent-knowledge/SKILL.md) | Knowledge | Local LLM-Wiki + cross-agent sharing + cross-session memory | ## Quick Start # 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. ## Design Principles - **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 ## Source Extracted from [SecAgent-TUI](https://github.com/JesusConwellpy/SecAgent-TUI) — a deep fork of DeepSeek-TUI built as a multi-agent security research platform. ## License MIT