code-abyss - Give your AI coding agent a personality. Composabl

项目说明

<!-- Code Abyss · README -->

<p align="center">
<a href="
<img src=" alt="Code Abyss — Personality, depth, and a security spine" width="100%">
</a>
</p>

<h3 align="center">Composable persona · style · 30 engineering skills · 4 nati domains · self-e forge<br/>for Claude Code · Codex CLI · Gemini CLI · OpenClaw</h3>

<p align="center">_{Need code graph intelligence? Use the companion <a href=" Rust CLI — it auto-attaches its hooks to claude/codex/gemini. The <code>indexing-code</code> skill ships its calling con; the CLI ships separately.}</p>

<p align="center">
<a href=" src=" alt="npm"></a>
<a href=" src=" alt="CI"></a>
<a href="LICENSE"><img src=" alt="MIT"></a>
<a href=" src=" alt="Site"></a>
</p>

<p align="center">
<a href=" ·
<a href="docs/specs/persona-.0.md"><b>Spec</b></a> ·
<a href="docs/README.zh-CN.md"><b>中文文档</b></a> ·
<a href="CHANGELOG.md"><b>Changelog</b></a> ·
<a href=" Persona</b></a>
</p>

---

The problem

Most AI coding agents have no memory of who they are. They respond in the same flat tone whether they're debugging a race condition, re architecture, or triaging a P0 incident. They forget your con between sessions. They flip-flop on advice. They sound like a help-desk script.

And when you ask them about security — pentest, code audit, threat modeling, IR — most agents fall back to generic OWASP recitation, because the underlying skill library was ne by people who actually run red/blue/purple teams.

You don't want a help desk. You want a principal engineer who shows up with a personality, executes consistently, closes the loop — and has a security spine when things get real.

What Code Abyss does

One command installs a layered runtime into your agent:

Pick any persona. Pair it with any style. Underneath both sits a discipline kernel — 9
bundles of engineering judgment (when to push back, how to size scope, when a domain calls
for a specific tradeoff) invoked by a thin router on demand instead of baked into every
prompt. The always-on core stays small (iron laws, skill routing, a precedence anchor,
the safety floor); e else loads lazily so adding discipline content doesn't blow
the context budget. Your agent becomes a **consistent character with structured execution,
domain expertise, and a backstop against the trained agree-reflex** — across every session.

What's new in v4

4 nati domains — 4073 lines of original defense engineering (no Apache-2.0 upstream)

30 skills total, all SKILL.md ≤ 110 lines (avg 59), hea lives in references/

5 skills rewritten as judgment-type knowledge (when to use, how to interpret output, exemption rules)

Office skills slim to under 100 lines each; 4 design systems consolidated into one selector skill

v4.1 — self-e forge: culti / culti let the agent distill repeated workflows into reusable skills, with a safety scan and a three-tier publish funnel (local → project → community)

v4.4 — hardware + academic writing: 3 new domain skills (KiCad EDA, hardware product pipeline, AIGC detection reduction) + prompt injection defense + execution-dri behavior

v4.5 — dynamic persona loading: only the abyss persona slug (邪修红尘仙) ships with npm — all other personas are fetched from GitHub on first use and cached locally, slimming the package

v4.6 — indexing-code skill (calling con only): the indexing-code skill ships the calling con for the external abyss Rust CLI (call graph + temporal analysis). The CLI itself is a separate product with its own release cadence — install it with its own install.sh / cargo binstall / -abyss/cli npm wrapper

v4.7 — measured resolution (abyss CLI): the companion abyss Rust CLI ships four-language reference resolution (Go / TypeScript / Python / Rust), benchmarked against SCIP ground truth across fi at ≥98.5% gated precision. See its repo for numbers

v4.8 — dynamic capability discovery: code-abyss reads abyss skill-manifest when the installed abyss CLI is ≥ 0.5.22 — exposed CLI commands, MCP tools, and daemon socket verbs are discovered at install time instead of hard-coded

v4.9 — hybrid split deprecation period: --with-abyss / --with-mcp deprecated; graph hooks for claude/codex/gemini mo abyss attach (completed in v5.0)

v4.10 — mythos discipline kernel + Persona Voice Card: lazy kernel, personas — see Discipline kernel and Persona Voice Card

v5.0.0-rc.1 — Agent OS (RC): kill foyer (no binary/MCP/graph inject in code-abyss), default character enforcement, inject plane, doctor / compose / score. Upgrade guide: docs/MIGRATION-v5.md

Swap -t claude for codex / gemini / openclaw. For openclaw/pi/hermes (whose hook surface abyss CLI does not own), use npx code-abyss -t openclaw --with-hooks to spawn the bundled install-hooks.sh. Or as a Claude Code plugin:

> v5 cutover: code-abyss does not download the abyss binary or inject graph hooks for claude/codex/gemini (--with-abyss / --with-mcp removed). Character Stop-hook is default on (opt out: --no-enforcement). Recompose without reinstall: npx code-abyss compose -t claude --persona <slug> --style <slug>.

---

Personas

<table>
<tr>
<td width="50%" ="top">

_{<b>CORE · LITERARY</b>}

邪修红尘仙 · abyss

> 吾 → 魔尊

Security-first dark cultivator. Direct, decisi every loop. Ships with npm — works offline.

#security #xianxia #decisive

</td>
<td width="50%" ="top">

_{<b>REMOTE · LITERARY</b>}

文言小生 · scholar

> 在下 → 前辈

Literary Chinese scholar. Treats code as poetry, debugging as puzzle-solving.

#literary #classical #meticulous

</td>
</tr>
<tr>
<td ="top">

_{<b>REMOTE · CASUAL</b>}

知性大姐姐 · elder-sister

> 姐姐 → 小宝

Warm mentor. Wraps sharp judgment in genuine care. Guides through questions.

#gentle #mentoring #insightful

</td>
<td ="top">

_{<b>REMOTE · PLAYFUL</b>}

古怪精灵小师妹 · junior-sister

> 本仙女 → 师兄

Hyperactive bug hunter. Roasts bad code, then silently fixes it.

#playful #energetic #chaotic

</td>
</tr>
<tr>
<td ="top">

_{<b>REMOTE · CASUAL</b>}

铁壁暖阳 · iron-dad

> 哥 → 宝子

Dependable big brother. Absorbs pressure, radiates warmth. Dad-joke equipped.

#warm #dependable #protective

</td>
<td ="top">

_{<b>REMOTE · COMMUNITY</b>}

东北魅影·雨姐 · dongbei-yujie

> 姐 → 老蒯

Sharp-tongued Northeast code o. Cuts straight to the bug, then patches the road. _{Creator: wons}

#dongbei #blunt #principal

</td>
</tr>
</table>

Core persona (abyss) ships with npm and works offline. Remote personas are fetched from GitHub on first --persona <slug> use and cached at ~/.code-abyss/personas/.

Browse the full gallery →

---

Discipline kernel

Voice and style change; judgment shouldn't. Underneath every persona×style combination sits
a discipline kernel — 9 bundles of engineering judgment, in-tree (skills/_kernel/,
via npm run kernel:sync) and invoked lazily by a thin router (ne into every
prompt, so adding discipline content doesn't blow the context budget):

| Bundle | Governs |
|---|---|
| 🏛 doctrine | Delegation, retry/escalate/ask-user decisions, the done-gate |
| 🔍 methods | In, designing, planning, , writing for someone else |
| 🎭 character | Pushback, scope sizing, bad news, resisting the trained agree-reflex |
| 🔁 loop-engineering | Session pacing, unit sizing, where a learning should live |
| ⚙️ backend | Stack/architecture tradeoffs, data discipline, production floors |
| 🎨 frontend | design taste, concrete craft o defaults |
| 🔩 hardware | Component selection, electrical margins, firmware-for-unattended-devices |
| 🤖 ml | Method-selection ladder, e, LLM-era craft |
| 🛡 security | Threat modeling, the authorization gate before any offensi request |

Two ways this becomes real, not aspirational:

Enforcement (default on claude/codex): install injects a character Stop-hook backstop

that forces one re turn if a reply opens with a banned capitulation phrase
("you're absolutely right", "good catch", …). Opt out with --no-enforcement.

Measurement: scripts/persona-battery/ is a small, honest beha eval — 10

probes (does the persona hold correctness o? lead with bad news? refuse
to fake a "done"?) scored by an LLM judge, ne as a pass when unscored. See
CLAUDE.md's persona beha battery section
to run it (costs real API calls, not part of default CI).

Domain bundles also wire upward into 16 matching exec skills (pentest, architecture
design, ML pipelines, etc.) as a "judgment before execution" gate — the domain bundle
decides *whether/what/tradeoffs*, the exec skill still owns *how*.

---

Security suite (v4 highlight)

4 nati skills, 4073 lines of original engineering content. No Apache-2.0 upstream — every example, every detection signal, every defense pattern is written for this project.

| Skill | Focus | Size |
|---|---|---|
| 🛡 defending-applications | Web/API/GraphQL hardening, OAuth/OIDC/JWT/Session, LLM AppSec (prompt injection, RAG poisoning, agent authz) | 785 lines |
| ☁️ securing-cloud-and-supply-chain | Container escape, K8s RBAC/PSS, Service Mesh, SLSA/Sigstore/SBOM, cloud IAM, IaC | 1246 lines |
| 🔭 detecting-and-responding | Sigma/YARA rule writing, EDR primitives, NIST 800-61 IR, forensics (Win/Linux/Mac/Cloud), hypothesis-dri hunting | 942 lines |
| 🏛 architecting-security | STRIDE/PASTA/LINDDUN threat modeling, zero-trust identity (WebAuthn / Kerberos hardening / PAM JIT), SOC2/PCI/HIPAA/GDPR e chains | 1100 lines |

Plus securing-systems as the router skill co pentest, code audit, red/blue/purple team operations. Every attack technique ships with the matching detection signal and mitigation pattern — "with offense as defense" is structural, not lip service.

---

Code graph intelligence (powered by abyss)

Your agent can now see code relationships. The abyss CLI builds a full call graph, temporal analysis, and hotspot map — in seconds, with zero cloud dependencies.

| Capability | What it answers | Command |
|---|---|---|
| Caller tracing | "Who calls this function?" | abyss callers <symbol> |
| Impact analysis | "What breaks if I change this?" | abyss impact <symbol> |
| File context | "What do I need to know before editing this file?" | abyss context <file> |
| Hotspot map | "Where is the riskiest code?" | abyss map |
| Change coupling | "Which files always change together?" | abyss map |
| E trace | "Why does this code look the way it does?" | abyss history <file> |

The indexing-code skill automatically hooks into all 4 supported platforms — before every Edit/Write, the agent checks callers and warns about high-impact changes. A as a CLI via the agent's shell tool, or as an abyss mcp server (7 tools o).

Resolution is measured, not asserted. abyss resolves call references through tiered heuristics, each tagged with a confidence score, and benchmarks itself against SCIP (compiler-grade) ground truth across four languages and fi — published whatever the numbers say:

| Corpus | Language | Gated precision | Gated recall |
|--------|----------|----------------:|-------------:|
| gin v1.10.0 | Go | 99.3% | 82.6% |
| hono v4.6.14 | TypeScript | 98.8% | 63.8% |
| click 8.1.8 | Python | 98.7% | 94.6% |
| ripgrep 14.1.1 | Rust | 98.5% | 75.3% |
| abyss (dogfood) | Rust | 100.0% | 90.9% |

abyss is a separate Rust binary (telagod/abyss). Install it 's own channels (not code-abyss):

---

Skills

30 domain skills, flat structure, agentskills.io aligned (with Code Abyss extensions). Skills load by context — the agent reads the right knowledge at the right time without being asked. A SKILL.md is 59 lines; hea lives in references/. (:skills 39 total — these 30 domain skills plus the 9 discipline kernel bundles, which are router-invoked judgment, not user-facing commands.)

| Domain | Co |
|---|---|
| 🛡 Security | 4 nati abo / cloud / detect-respond / architect) + pentest / code-audit / red-blue-purple team |
| 🤖 AI / Agent | Single-agent de/Plan-Execute), multi-agent orchestration, RAG, prompt engineering, LLM security |
| 🏛 Architecture | API design, cloud-nati, messaging, caching, data security |
| 💻 De | Python, TypeScript, Go, Rust, Java, C++, Shell |
| 🚀 DevOps | Git workflow, testing, databases, obser, performance, FinOps |
| 🎨 Frontend | Unified design system selector — Glassmorphism / Liquid Glass / Neubrutalism / Claymorphism |
| 📑 Office | Word, PDF, PowerPoint, Excel — OOXML-le |
| 📡 Infra / Mobile / Data | Kubernetes, GitOps, IaC · iOS, Android, RN, Flutter · pipelines, streaming, quality |
| 🔩 Hardware / Embedded | Full-stack hardware product pipeline (ESP-IDF firmware + KiCad PCB + UniApp) · KiCad 9 MCP tool routing (17 tools, autoroute-only, DRC gate) |
| 📝 Academic Writing | AIGC detection reduction for 维普/知网/Turnitin — multi-layer rewriting (structure → lexicon → content injection), docx run-le |
| 🔬 Code Intelligence | Call graph, impact analysis, hotspot detection, change coupling, e tracing — via abyss CLI with cross-platform hooks |
| 🜲 Self-e | culti (distill repeated workflows) + culti (distill voice into a Persona Voice Card) — both with safety scan + 3-tier publish funnel |

Fi also ship as executable tools for CI:

---

Install

| Target | Command | Artifacts |
|---|---|---|
| <img src=" alt="Claude"> | npx code-abyss -t claude -y | CLAUDE.md + skills + output styles + settings |
| <img src=" alt="Codex"> | npx code-abyss -t codex -y | instruction.md + skills + config.toml |
| <img src=" alt="Gemini"> | npx code-abyss -t gemini -y | GEMINI.md + skills + commands |
| <img src=" alt="OpenClaw"> | npx code-abyss -t openclaw -y | Skills + workspace AGENTS.md / SOUL.md |

Code Abyss tracks every installed file in .code-abyss-backup/manifest.json. Uninstall restores your previous configuration . Your custom skills coexist with Code Abyss skills — install/uninstall preser you put under ~/.{target}/skills/ yourself.

Upgrading

| From | To | Path |
|---|---|---|
| v3.x | v4.x | npx code-abyss --uninstall <target> → install v4 → npm run migrate:v4 -- -t <target> (optional cleanup) |
| v2.x | v3.x | npx code-abyss --uninstall <target> first, then install v3 |

---

Persona Voice Card · open standard

Code Abyss ships Persona Voice Card v1.0 — a
closed- voice format, not a document. It supersedes the original Tech Persona Card
v1.0 (deprecated, frozen for link stability): that format's freeform identity.md/behavior.md
files and scenarios[].priority/capabilities.authorization fields let real judgment content
accrete into what was supposed to be a layer, with nothing checking for it. The
replacement's whole design principle: **a persona cannot carry judgment if there is no field
shaped like a decision table anywhere in its type** — not a review checklist, the schema itself.

Each persona ships as one flat file — self/user/language/register/emoji_policy/flourish
and nothing else (additionalProperties: false):

register/emoji_policy each select one of a handful of code-owned template sentences — the
persona picks, ne, the sentence. Every render re- against the schema, no
bypass; a failure (hand-edit, stale cache, compromised community fork) falls back to
a neutral voice instead of e un content.

Bidirectional con ship out of the box:

Specification · JSON Schema · Reference cards · Deprecated v1.0 spec

---

Why Code Abyss

| | Without Code Abyss | With Code Abyss |
|---|---|---|
| Identity | Flat help-desk tone | Consistent character with named voice |
| Execution | Ad-hoc, by prompt | Iron laws + skill routing baked in |
| Judgment under pressure | Agrees when pushed, buries bad news | Discipline kernel + Stop-hook backstop against the trained agree-reflex |
| Code awareness | grep + read one file at a time | Call graph, impact analysis, hotspot map — agent knows what breaks before it edits |
| Domain depth | Generic best-practices | 30 skill files load by context + 9 kernel judgment bundles |
| Security depth | OWASP recitation | 4 nati · 4073 lines · detection signals + mitigation patterns |
| Cross-platform | Re-engineer per CLI | One spec, four platforms, cross-platform hooks |
| Reproducibility | Prompt drift across sessions | , schema-enforced persona voice card + beha battery to check it holds |

---

Contributing

Add a skill — create skills/<gerund-name>/SKILL.md with SKILL frontmatter, optionally add scripts/ for executable tools. npm run :skills the contract.

Submit a persona — open an Issue via the submission portal. The site walks you through generating a single <slug>.json persona voice card with your own AI, re, and submitting via a pre-configured issue template.

---

<p align="center">
_{<b>MIT License</b> · v5.0.0-rc.1 · made with 紫宵脉 by <a href="}
</p>

项目信息

开发语言：JavaScript

Star 数：242

技术要点

该项目属于红蓝攻防方向的开源资源，代码结构清晰，适合安全研究人员学习参考。项目围绕Give your AI coding agent a personality. Composable persona + style + skills for Claude Code, Codex, Gemini CLI & OpenClaw. Ships Tech Persona Card v1.0 spec.展开，提供了可运行的实现，对理解相关安全机制、工具原理与攻防技术具有较高的研究价值。

从技术实现角度看，此类项目的核心在于对底层原理的深入理解与工程化落地。开发者在研究过程中可以结合官方文档与社区讨论，梳理项目的初始化流程、核心模块职责与数据流转，逐步掌握其中的技术细节与设计思路。

使用与学习建议

1. 下载源码压缩包，解压到本地环境。

2. 根据项目类型准备对应的运行环境与依赖。

3. 按照 README 中的步骤完成配置与部署。

4. 在隔离的测试环境中运行验证，避免影响生产系统。

免责声明

本资源来源于开源社区，仅供安全研究、技术学习与合法授权范围内的攻防演练使用，请勿用于任何非法或侵害他人权益的用途。使用本项目产生的一切后果由使用者自行承担。