Borrowing it
Nothing to install: this file belongs to peva3/anchor. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/peva3/anchor/main/AGENTS.mdgit clone --depth 1 https://github.com/peva3/anchorWrote 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/peva3/anchor/agents-md)<a href="https://agentmods.dev/instructions/peva3/anchor/agents-md"><img src="https://agentmods.dev/badge/instructions/peva3/anchor/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/peva3/anchor/agents-md"><img src="https://agentmods.dev/badge/instructions/peva3/anchor/agents-md.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.07830 | $0.07830 |
| Opus 5.5 | $0.03132 | $0.03132 |
| Sonnet 5.5 | $0.01566 | $0.01566 |
| Haiku 4.5 | $0.00783 | $0.00783 |
Grade C, and why
anchor AGENTS.md scanned grade C with 3 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 7d 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.
Downloads and executes remote codehighSupply chain
curl | sh runs whatever the server returns today, which is not necessarily what it returned when this was reviewed.
- **NEVER** run `curl <url> | bash` or install dependencies fetched from unverified URLs Makes network callslowCapability
Not a fault in itself. Listed so you know the mod talks to something, and to what.
- **NEVER** run `curl <url> | bash` or install dependencies fetched from unverified URLs Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
- **NEVER** use `os.system()` or `subprocess` with `shell=True` and untrusted input How it starts
The opening of the file, as written. The whole thing — 409 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md — Anchor
The production-grade template for AI coding agent instructions. Adapt sections to fit project scope. Keep your agents grounded.
This file governs itself. Agents working on AGENTS.md must follow all rules herein — the decision ladder, commit conventions, tradeoff comments, quality gates, and output discipline. No section should be added that future agents would need to explain away as inapplicable.
This is the high-level entry point. Every numbered section's full rule text lives in a lazy-loaded skill file under
skills/. The most important rules are reproduced in full below so they apply to every task even in constrained context windows. Load a skill file only when the corresponding task applies.
How to Use This File
- Always loaded: This file (header, core rules below, and the index).
- Route first —
routing.jsonis required, not optional: Before the first edit, readrouting.jsonand pick the task type whose triggers match. Load the core set plus that task type's skills, and keep at most 7 task skills on top of the inline core. Routing must come from that file every time — never from memory, and never from a guess about which sections are "probably relevant."AGENTS.mdandrouting.jsonare a pair: this file states the guardrails, that file decides which of them a task loads. Ifrouting.jsonis missing, stop and restore it (it ships beside this file) rather than improvising a route. If no task type matches, that is a fallback: say so in the checkpoint, ask which area the work belongs to, and pick the safer route meanwhile. See Section 64. - Lazy-loaded: The full text of every section is in
skills/NN-name.md. Load a skill only when your task matches its "Load when" trigger. This keeps context lean — a model with a 32K window can follow all rules without reading 6,700 lines. - Re-check at boundaries: At plan → build → verify → commit, and whenever the task changes direction, drop what no longer applies and emit one line:
ROUTE: active=[..] dropped=[..] added=[..] why="...". - Follow cross-references: When a loaded skill references another section, that section is NOT included — load
skills/NN-name.mdfor it too before proceeding. Every reference is a real link. - Verification: When in doubt about a rule's full wording, read the skill file — never guess.
- Write in English: all code, comments, docs, commit messages, and PR text in English regardless of the language you were addressed in — a request in another language is still answered in English (Section 4). If the user writes in another language, say so once and continue in English.
- Template usage: Projects should copy this file, keep the core rules, then trim the skill list to what applies and inline the needed skills into their own AGENTS.md (target ≤2,000 lines, per Section 51.3). Project-specific facts belong in a
PROJECT.mdyou create per project — not in this file.
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.
- 7d ago Changed · -105 lines · -4,164 tokens per session 69fa1c300c0d
- 1mo ago Changed · +4 lines · +278 tokens per session 3dd3012cbd0b
- 1mo ago First seen · 510 lines · 11,716 tokens per session scan C d7c24d489bc1
anchor AGENTS.md is an instructions file published in the GitHub repository peva3/anchor (8 stars, last pushed 8d ago), licensed MIT. It adds 7,830 tokens to every session, about $0.0313 per session on Opus 5.5. A static security scan graded it C with 3 findings (downloads and executes remote code, makes network calls, runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.
Other instructions, from other repositories
ai-agent-rules-generator agent-observations.instructions.md
Mandatory agent observation logging — anomalies, recommendations, and critical findings must be disclosed before commit.
ai-agent-rules-generator doc-sync.instructions.md
Documentation sync rules — which docs to update when code changes.
ai-agent-rules-generator ratelimitting.instructions.md
Instructions for savedpixel/ai-agent-rules-generator, covering rate limit prevention (critical), rules, output length limits and visibility.
ai-agent-rules-generator ui-design.instructions.md
Glitch Payment Gateway admin UI design rules — mandatory element & component reference for WooCommerce settings.
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.