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/shisa-ai/textguard/agents-mdgit clone --depth 1 https://github.com/shisa-ai/textguardWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/instructions/shisa-ai/textguard/agents-md)<a href="https://agentmods.dev/instructions/shisa-ai/textguard/agents-md"><img src="https://agentmods.dev/badge/instructions/shisa-ai/textguard/agents-md.svg" alt="Measured on agentmods" height="20"></a>What 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 | $0.01453 | $0.01453 |
| Opus 5 | $0.00727 | $0.00727 |
| Sonnet 5 | $0.00291 | $0.00291 |
| Haiku 4.5 | $0.00145 | $0.00145 |
Grade A, and why
textguard AGENTS.md scanned grade A with 0 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 4d 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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 115 lines — stays where its author put it; the contents beside it link to each section on GitHub.
textguard Agent Guide
textguard is a greenfield Python package being extracted from shisad work on hostile-text normalization and detection. Keep the security ideas strong, keep the package reusable, and avoid importing daemon/framework complexity that does not belong in a standalone library.
Instruction precedence: if this file conflicts with platform, system, or developer instructions, follow those first.
Current Source Of Truth
README.mdis the public package overview and usage surface.docs/PLAN.mdis the working implementation plan.WORKLOG.mdis the chronological development log. Append entries when making design decisions, completing work, or changing direction. Do not rewrite earlier entries — the line-number stability matters for referencing.CHANGELOG.mdis the user-facing release history, reverse chronological. Update only when shipping a release or pre-release milestone.- Use
shisadas a reference source for reusable text-defense logic, not as an architecture template.
First Principles
- Preserve legitimate text by default. Normal multilingual Unicode input is a first-class use case, not collateral damage.
- Do not make "convert everything to ASCII" the default behavior. Lossy or destructive transforms must be explicit opt-in.
- Security features must improve safety without turning the package into a blunt text shredder.
- Keep the package focused on text normalization, decoding, inspection, and sanitization primitives for LLM-adjacent systems.
- Prefer pure, testable library code over framework coupling, background services, policy engines, or repo-specific runtime assumptions.
- Bound all decoding and expansion behavior. No unbounded recursion, no unbounded size blowups, no regexes with obvious pathological behavior.
Intended Project Shape
Build toward a small public-package layout like this:
textguard/
|- src/textguard/ # library code
|- tests/ # unit and integration tests
|- pyproject.toml # packaging metadata
|- README.md # public install and usage docs
|- docs/PLAN.md # optional implementation plan for non-trivial work
`- docs/PUBLISH.md # optional release checklist
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.
- 4d ago First seen · 115 lines · 1,453 tokens per session scan A 0e8f3a690276
textguard AGENTS.md is an instructions file published in the GitHub repository shisa-ai/textguard (2 stars, last pushed 4mo ago), licensed Apache-2.0. It adds 1,453 tokens to every session, about $0.0073 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
aigis CLAUDE.md
Claude Code instructions for killertcell428/aigis, covering aigis — project guidelines, zenn記事管理（zenn cli + github連携）, ディレクトリ構成, 記事の作成 and プレビュー.
SecureAI-Scan CLAUDE.md
Instructions for akanthed/SecureAI-Scan, covering claude.md, what this is, hard requirements (non-negotiable, apply to every change), 1. zero tolerance for false positives and commands.
iris CLAUDE.md
Claude Code instructions for rux-eth/iris, covering claude.md, what is this project, build & test commands, architecture and implementation status.
guidecheck CLAUDE.md
Claude Code instructions for snapsynapse/guidecheck, covering claude.md, purpose, tech stack, directory layout and conventions.
ai-dial-core CLAUDE.md
Claude Code instructions for epam/ai-dial-core, covering claude.md, build & run, set credentials via environment variables, build (skip tests) and run all tests.
ai AGENTS.md
AGENTS.md instructions for vercel/ai, covering agents.md, project overview, repository structure, key directories and core package dependencies.