菜单
Use when a MASM procedure needs temporary scratch storage — keep it in procedure-local memory so it cannot collide in a shared global region.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/0xMiden/agent-tools --skill masm-locals-over-globals命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Use when writing kernel, account, or note MASM code that reads from or writes to the advice provider (advice stack / advice map) — validate advice data.
Use when writing or reviewing MASM hot paths — prefer the cheaper equivalent instruction: `neq.0` over `push.0 gt` for non-zero checks, `cdrop` over an `if/else` selecting between two values, `dup.N` over `loc_load` for a value still on the stack, `eqw` over hand-rolled element-wise word comparison, `u32gt`/`u32lt` over generic `gt`/`lt` on known-u32 operands.
Use when writing a GENERIC MASM storage utility (one that operates over a caller-chosen slot or map) inside a reusable account component — receive the slot id as a parameter so the utility works against any slot, not one hard-coded one.
基于 SOC 职业分类
正在显示 SKILL.md
| name | masm-locals-over-globals |
| description | Use when a MASM procedure needs temporary scratch storage — keep it in procedure-local memory so it cannot collide in a shared global region. |
When a MASM procedure needs scratch storage that lives only for the duration of one invocation, use procedure-local memory rather than allocating in a shared global memory region.
Declare how many locals the procedure needs with @locals(N) directly above the procedure, then read and write them:
loc_store / loc_load for a single element.loc_storew_le / loc_loadw_le (or the _be variants) for a full word.The bare loc_storew / loc_loadw instructions have been removed and now raise a deprecation error — use the _le / _be word variants instead.
Global memory regions are reserved for state that crosses procedure boundaries (kernel inputs, account data, advice-keyed state). Stashing per-call scratch there leaks an implementation detail into a shared namespace and ties the procedure to a fixed address.
Procedure locals are addressed relative to the procedure's frame pointer (FMP) and bounded by its declared @locals(N) count: the assembler emits a prologue that advances the FMP to allocate the frame on entry and an epilogue that restores it on exit, so two callers of the same procedure can't collide. A hard-coded scratch slot in global memory writes to a fixed absolute address with no per-procedure isolation, so it risks colliding with another procedure using the same address, forces every caller to avoid clobbering it, and locks the layout.
A procedure with no @locals declaration allocates zero locals, so any loc_store / loc_load in it fails to assemble — always declare the locals you use.
# Good
@locals(2)
proc compute_hash
# write the two scratch values into local slots 0 and 1
loc_store.0
loc_store.1
# ...
loc_load.0
loc_load.1
end
# Bad: scratch in a shared global region
const SCRATCH_PTR = 0x4000
proc compute_hash
mem_store.SCRATCH_PTR # collides with anyone else using SCRATCH_PTR
end