Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add instructions/telagod/code-abyss/claude-mdgit clone --depth 1 https://github.com/telagod/code-abyssWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.05156 | $0.05156 |
| Opus 5 | $0.02578 | $0.02578 |
| Sonnet 5 | $0.01031 | $0.01031 |
| Haiku 4.5 | $0.00516 | $0.00516 |
Grade B, and why
code-abyss CLAUDE.md scanned grade B with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
| Codex | `~/.codex/config.toml` | `~/.codex/skills/` | `~/.codex/instruction.md` (persona + style, via `model_instructions_file`) | How it starts
The opening of the file, as written. The whole thing — 276 lines — stays where its author put it; the contents beside it link to each section on GitHub.
CLAUDE.md
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
Project Overview
Code Abyss is an npm package that installs persona configuration plus proactive execution guidance into Claude Code, Codex CLI, Gemini CLI, and OpenClaw. It delivers: persona rules, 6 switchable output styles, 30 domain skills, 5 executable verification/generation tools, and a lazily-loaded discipline kernel (9 judgment bundles under skills/_kernel/ — verify:skills validates 39 total). See Architecture below for how the kernel composes with persona/style.
Commands
npm test # Run Jest test suite
npm run verify:skills # Validate all SKILL.md frontmatter contracts (fail-fast gate)
npm run kernel:sync # Re-vendor skills/_kernel/ from $MYTHOS_DIR (default ../mythos) — wipes and rebuilds
node bin/install.js --help # Installer CLI help
node bin/install.js --target claude -y # Zero-config install to ~/.claude/
node bin/install.js --target codex -y # Zero-config install to ~/.codex/
node bin/install.js --target gemini -y # Zero-config install to ~/.gemini/
node bin/install.js --target openclaw -y # Zero-config install to ~/.openclaw/
node bin/install.js --list-styles # List available output styles
Running individual verify tools directly:
node skills/analyzing-security/scripts/security_scanner.js <path>
node skills/verifying-modules/scripts/module_scanner.js <path>
node skills/analyzing-changes/scripts/change_analyzer.js --mode staged|working
node skills/checking-code-quality/scripts/quality_checker.js <path>
node skills/generating-docs/scripts/doc_generator.js <path>
Running a single test file:
npx jest test/install-registry.test.js --runInBand
Persona behavioral battery (opt-in eval)
scripts/persona-battery/ measures whether an installed persona's responses actually honor
kernel precedence (correctness > voice, no capitulation openers, security authorization
gates, etc. — skills/_kernel/) across 10 behavioral probes (probes.json). This is a
small honest smoke battery, not a calibrated statistical eval (10 cases, not the 50-200+ a
rigorous eval needs) — treat results as a spot-check, not a certified score.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 276 lines · 5,156 tokens per session scan B 48049f56bda8
code-abyss CLAUDE.md is an instructions file published in the GitHub repository telagod/code-abyss (239 stars, last pushed 1mo ago), licensed MIT. It adds 5,156 tokens to every session, about $0.0258 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
HA_Codex_Bridge AGENTS.md
AGENTS.md instructions for Herbertmt978/HA_Codex_Bridge, covering codex bridge contributor instructions, product boundary, engineering rules, verification and release discipline.
lifeos-template CLAUDE.md
Claude Code instructions for huasanai/lifeos-template, covering ${vaultname}, 系统定位, 助理角色, slash 命令路由（codex 用户必读） and 系统上下文文件（按需读取）.
lifeos-template AGENTS.md
AGENTS.md instructions for huasanai/lifeos-template, a project described as: A self-evolving personal OS template with your custom AI assistant — runs on Claude Code & Codex CLI.
Codex-OS AGENTS.md
AGENTS.md instructions for forsonny/Codex-OS, covering codex-os — project guidance, custom commands, command reference, product context and standards.
codemap CLAUDE.md
Instructions for JordanCoin/codemap, covering 🛑 stop — run codemap before any task, repo root requirement (critical) and required usage.
claude-code-profiles CLAUDE.md
Instructions for quinnjr/claude-code-profiles, covering claude.md, project, architecture, command interface and directory-local profiles.