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/lhl/realitycheck/agents-mdgit clone --depth 1 https://github.com/lhl/realitycheckWhat 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 | $0.03181 | $0.03181 |
| Opus 5 | $0.01590 | $0.01590 |
| Sonnet 5 | $0.00636 | $0.00636 |
| Haiku 4.5 | $0.00318 | $0.00318 |
Grade D, and why
realitycheck AGENTS.md scanned grade D with 2 findings 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.
Unrestricted tool accessmediumExcessive agency
A wildcard tool grant or "run any command" leaves no least-privilege boundary at all.
- NEVER run any command that can delete/overwrite existing work unless the user explicitly instructs it. Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
- `rm -rf ...` / overwriting redirects like `> file` How it starts
The opening of the file, as written. The whole thing — 342 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Reality Check - Development Guide
Please refer to README.md, docs/PLAN-separation.md, docs/IMPLEMENTATION.md, and docs/DEPLOY.md for project-specific details.
This AGENTS.md/CLAUDE.md is specifically for ground rules, process, and behavior notes.
Shared Repo / Multi-Agent Safety (MUST FOLLOW)
This repo may be edited by multiple agents concurrently. Treat the working tree as shared state.
Before any change (required)
- Run
git status --porcelain. - If the working tree has unrelated changes, proceed normally but do not touch them; stage only your own files (
git add <paths>/git add -p). - If you need to edit a file that already has changes you did not create (potential conflict), flag it and ask the user how to proceed.
- If there are merge conflicts, STOP and ask the user how to proceed.
Never discard others’ work (hard rule)
- NEVER run any command that can delete/overwrite existing work unless the user explicitly instructs it.
This includes (but is not limited to):
git restore ...git checkout -- ...git checkout .git reset --hard ...git clean -fd ...rm -rf .../ overwriting redirects like> file- bulk rewrites that destroy local edits (e.g., aggressive formatters) unless requested
Scope discipline
- Only edit files needed for the user’s current request.
- Do not “clean up”, refactor, or revert unrelated diffs in touched files.
- If you need to avoid committing unrelated diffs, use
git add <paths>orgit add -p; never “fix” by reverting other hunks.
If conflicts are unavoidable
- Coordinate: propose a plan that preserves both sets of edits (e.g., separate commits/branches/patches), and wait for explicit user instruction before any destructive resolution.
Project Overview
Reality Check is a framework for rigorous, systematic analysis of claims, sources, predictions, and argument chains. It provides:
- LanceDB-backed storage with semantic search
- Structured methodology for claim extraction and evaluation
- Evidence hierarchy and prediction tracking
- Claude Code plugin for workflow automation
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 · 342 lines · 3,181 tokens per session scan D bc1b0a0b2abe
realitycheck AGENTS.md is an instructions file published in the GitHub repository lhl/realitycheck (41 stars, last pushed 2mo ago), licensed Apache-2.0. It adds 3,181 tokens to every session, about $0.0159 per session on Opus 5. A static security scan graded it D with 2 findings (unrestricted tool access, recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
intelligent-terminal rust.instructions.md
Concise Rust coding conventions for this repository.
InvestSkill GEMINI.md
Instructions for yennanliu/InvestSkill, covering investskill — gemini cli setup & usage guide, installation & setup, quick start, navigate to the investskill directory and start gemini cli (loads gemini.md automatically).
she-love-me CLAUDE.md
Instructions for 863401402/she-love-me, covering claude.md and 唯一工作流.
wayland-core copilot-instructions.md
Instructions for FerroxLabs/wayland-core, covering ijfw rules, output discipline, memory routing, context discipline and cross-audit.
ZipAgent AGENTS.md
Instructions for JiayuXu0/ZipAgent, covering repository guidelines, project structure & module organization, build, test & development commands, coding style & naming conventions and testing guidelines.
agentconfig.org AGENTS.md
Instructions for agentconfig/agentconfig.org, covering agent instructions for agentconfig.org, project overview, target audience, site structure and tech stack.