tactics-deep

tactics-deep is an agent for Claude Code from Verified-zkEVM/evm-asm. It costs 0 tokens per session (3,559 once invoked), scanned A, original, MIT.

A project-specific guide to advanced tactics and architecture for EvmAsm, including low-level code automation, calling conventions, formal proofs, and Compose files. It is loaded only when those areas need work.

In plain words
What is it for?
Use it when changing frame-automation code, designing callable shims, writing three-level opcode proofs, splitting Compose files, or reading benchmark history.
Why use it?
It keeps the main project instructions short while providing the detailed rules needed for specialized changes.

Agent for Claude Code

Written for Claude Code: a Claude Code subagent (agents/*.md). Also seen: mentions AGENTS.md.

Install

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.

agentmods
npx agentmods add agents/verified-zkevm/evm-asm/tactics-deep
Clone the repo
git clone --depth 1 https://github.com/Verified-zkEVM/evm-asm

Made for: Claude Code.

Wrote 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.

agentmods badge for tactics-deep

README.md
[![agentmods](https://agentmods.dev/badge/agents/verified-zkevm/evm-asm/tactics-deep.svg)](https://agentmods.dev/agents/verified-zkevm/evm-asm/tactics-deep)
Your own site
<a href="https://agentmods.dev/agents/verified-zkevm/evm-asm/tactics-deep"><img src="https://agentmods.dev/badge/agents/verified-zkevm/evm-asm/tactics-deep.svg" alt="Measured on agentmods" height="20"></a>
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 3,559 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00000 $0.03559
Opus 5 $0.00000 $0.01780
Sonnet 5 $0.00000 $0.00712
Haiku 4.5 $0.00000 $0.00356

Measured 2d ago against content hash d57de74c0ec5, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-06, from the pricing page.

Security

Grade A, and why

tactics-deep 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 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.

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.

docs/agents/tactics-deep.md · 235 lines

How it starts

The opening of the file, as written. The whole thing — 235 lines — stays where its author put it; the contents beside it link to each section on GitHub.

EvmAsm — Deep Tactics & Architecture

Moved out of AGENTS.md to keep the agent guide compact. Load this when:

  • writing or restructuring frame-automation (runBlock, seqFrame, xperm, xcancel),
  • using the LP64 calling convention or designing a callable shim,
  • working on a three-level opcode proof (limb → composition → semantic stack spec),
  • splitting Compose files in parallel or working with the file-size guardrail,
  • consuming the benchmark-history orphan branch.

See also TACTICS.md (user-facing tactic reference) and GRIND.md (grindset definitions).

Frame-automation Tactics

Primary reference: TACTICS.md is the user guide for runBlock, seqFrame, xperm, xcancel, the @[spec_gen] registry, and the domain-specific grindsets (divmod_addr, rv64_addr, reg_ops, byte_alg). Read it before hand-writing a cpsTriple_seq_* chain or wiring a new @[...] equality-closing attribute from scratch.

Separation Conjunction Permutation Tactic

The sep_perm tactic (defined in SepLogic.lean) closes goals that require rearranging sepConj (**) chains. It works by AC-normalizing both the hypothesis and goal using simp with three equality lemmas:

  • sepConj_assoc' : ((P ** Q) ** R) = (P ** (Q ** R))
  • sepConj_comm' : (P ** Q) = (Q ** P)
  • sepConj_left_comm' : (P ** (Q ** R)) = (Q ** (P ** R))

Usage: Given a hypothesis h : (A ** B ** C) s and goal ⊢ (C ** A ** B) s:

sep_perm h

This handles arbitrary permutations of any number of assertions in a sepConj chain.

Additional equality lemmas for empAssertion elimination:

  • sepConj_emp_right' : (P ** empAssertion) = P
  • sepConj_emp_left' : (empAssertion ** P) = P

When rearranging involves memBufferIs (which unfolds to ... ** empAssertion), combine all rules in one simp:

simp only [memBufferIs, addr_100_plus_4, addr_104_plus_4,
  sepConj_emp_right', sepConj_emp_left',
  sepConj_assoc', sepConj_comm', sepConj_left_comm'] at hab ⊢
exact hab

Read the full file on GitHub · 235 lines

Changes

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.

  1. 2d ago Changed · +39 lines d57de74c0ec5
  2. 6d ago First seen · 196 lines · 0 tokens per session scan A afe4715ce260

Subscribe to this mod's changes

tactics-deep is an agent published in the GitHub repository Verified-zkEVM/evm-asm (55 stars, last pushed yesterday), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 3,559 tokens. 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 2026-08-30.