tomevault-io

masm-doc-comments

Enforce doc comment conventions for Miden Assembly (.masm) procedures. Use when editing, reviewing, or creating .masm procedures, especially when documenting inputs, outputs, panic conditions, or invocation types. Use when this capability is needed.

tomevault-io 1 Updated 1mo ago
GitHub

Install

npx skillscat add tomevault-io/tomes/masm-doc-comments

Install via the SkillsCat registry.

SKILL.md

MASM Procedure Doc Comments

Overview

Every public MASM procedure should have a doc comment block using #! prefix with these sections in order:

  1. Description - What the procedure does
  2. Inputs/Outputs - Stack state before/after
  3. Where - Explanation of each stack item (omit when the Description already names every item)
  4. Panics if - Error conditions (when applicable)
  5. Invocation - How the procedure is called (exec, call, or dyncall; omitted for syscall-invoked kernel procedures)

Required Format

#! Brief description of what the procedure does.
#!
#! Additional context if needed.
#!
#! Inputs:  [item1, item2, WORD_ITEM]
#! Outputs: [result1, RESULT_WORD]
#!
#! Where:
#! - item1 is the description of item1.
#! - item2 is the description of item2.
#! - WORD_ITEM is a 4-element word representing X.
#! - result1 is the description of result1.
#! - RESULT_WORD is the resulting word.
#!
#! Panics if:
#! - condition that causes the procedure to fail.
#! - another error condition.
#!
#! Invocation: exec
pub proc my_procedure

Stack Notation

Naming Conventions

Type Style Example
Single felt lowercase with underscores note_index, amount, balance
Word (4 felts) UPPERCASE with underscores ASSET, RECIPIENT, SCRIPT_ROOT
Multi-felt (2-3) lowercase with {parts} suffix account_id_{suffix,prefix}
All-zero Word EMPTY_WORD [..., EMPTY_WORD, ...]

EMPTY_WORD is a naming convention used in stack trackers, Where: bullets, and prose comments to denote the all-zero Word [0, 0, 0, 0].

In composite braces, list parts in stack-top-first order, no spaces inside the braces: account_id_{suffix,prefix} because the suffix sits on top of the stack and the prefix below it. The same rule applies to other split-128-bit IDs (sender_{suffix,prefix}, faucet_id_{suffix,prefix}, etc.).

Stack Order

Items are listed left-to-right, with the top of stack first:

#! Inputs:  [top_item, second_item, THIRD_WORD]

Empty Stack

Use empty brackets for no inputs or outputs:

#! Inputs:  []
#! Outputs: [result]

Span notation (N) family

(N) after an item name denotes a span of N felts (not a Word). Spans stay lowercase; Words stay UPPERCASE and never take (N). pad(N) (see masm-padding skill) is one member of this family; other spans appear in protocol code:

#! Inputs:  [first_element, foreign_procedure_inputs(15)]
#! Outputs: [foreign_procedure_outputs(16)]

Padding (for call and dyncall procedures)

Procedures entered at the stack-depth-16 floor (call and dyncall) must show explicit padding so Inputs and Outputs each sum to 16 elements (see masm-padding skill):

#! Inputs:  [ASSET, pad(12)]
#! Outputs: [pad(16)]
#!
#! Invocation: call

Where Section

Define every item from Inputs and Outputs:

#! Where:
#! - note_index is the index of the input note.
#! - sender_{suffix,prefix} are the suffix and prefix felts of the sender ID.
#! - ASSET_ID is the asset ID of the asset [0, 0, faucet_id_suffix, faucet_id_prefix].
#! - balance is the fungible asset balance in the vault.

Rules:

  • Use "is" for single items, "are" for multi-part items
  • Start descriptions lowercase (continues the sentence)
  • End each line with a period
  • Group related items (e.g., all inputs, then all outputs)
  • Avoid including low-level details, e.g. how a value is computed.
    • Good: NOTE_DETAILS_COMMITMENT is the commitment to the note's details.
    • Avoid: NOTE_DETAILS_COMMITMENT is the commitment to the note's details computed as hash(RECIPIENT_DIGEST || ASSETS_COMMITMENT).

When Where: may be omitted

Omit the Where: section entirely when the Description already names every Inputs/Outputs item and adding bullets would just repeat that information. Common case: trivial accessor procedures.

#! Returns the maximum supply.
#!
#! Inputs:  [pad(16)]
#! Outputs: [max_supply, pad(15)]
#!
#! Invocation: call
pub proc get_max_supply

No Where: is needed: the single named output max_supply is already identified by the Description. If you would otherwise write #! - max_supply is the maximum supply., skip it.

Add Where: whenever any item needs description beyond what the Description line conveys — different name, additional constraint, composition (ASSET_ID = [0, 0, faucet_id_suffix, faucet_id_prefix]), or anything non-obvious.

Panics Section

Bullets describe the condition, not the error identifier. Write the nonce has already been incremented., not ERR_ACCOUNT_NONCE_CAN_ONLY_BE_INCREMENTED_ONCE.. The identifier is an implementation detail of the assert; the doc comment should read as English.

Direct Panics

List conditions from assert* statements in the procedure:

#! Panics if:
#! - flag is false.
proc sample_procedure
    # => [flag]
    assert.err=ERR_FLAG_IS_FALSE

Propagated Panics

When calling other procedures that may panic:

Simple case (< 4 conditions): List the specific conditions:

#! Description, inputs, etc.
#! Panics if:
#! - flag is false.
proc sample_procedure
    # => [flag]
    exec.another_procedure # this procedure may panic
end

proc another_procedure
    # => [flag]
    assert.err=ERR_FLAG_IS_FALSE
end

Complex case (4+ conditions): Reference the subprocedure:

#! Description, inputs, etc.
#! Panics if:
#! - another_procedure fails to verify.
proc sample_procedure
    # => [flag_1, flag_2, flag_3, flag_4]
    exec.another_procedure # this procedure may panic
end

#! Description, inputs, etc.
#! Panics if:
#! - flag_1 is false.
#! - flag_2 is false.
#! - flag_3 is false.
#! - flag_4 is false.
proc another_procedure
    # => [flag_1, flag_2, flag_3, flag_4]
    assert.err=ERR_FLAG_1_IS_FALSE
    assert.err=ERR_FLAG_2_IS_FALSE
    assert.err=ERR_FLAG_3_IS_FALSE
    assert.err=ERR_FLAG_4_IS_FALSE
end

No Panics

Omit the "Panics if:" section entirely if the procedure cannot panic.

Invocation Types

Specify how the procedure should be invoked. The value matches the MASM instruction that user-code callers use to enter the procedure:

Value Used by callers as When to use
exec exec.<proc> Standard inline call. Shares the caller's stack; no padding requirement.
call call.<proc> Cross-context call (e.g. into another account). Enters at stack depth 16 — Inputs/Outputs must show pad(N) to total 16 (see masm-padding).
dyncall dyncall from a script Entry point of a note script or transaction script. Stack-depth-16 floor applies on entry.
#! Invocation: exec
#! Invocation: call
#! Invocation: dyncall

For existing procedures, pick the value that matches how callers invoke them: call when invoked via call.<procedure_name>, exec for exec.<procedure_name>, dyncall for note-script and transaction-script entry points.

Kernel procedures (syscall-invoked) are exempt

Kernel procedures under crates/miden-protocol/asm/kernels/ are invoked by the VM via syscall.<proc> from user code. They do not carry an Invocation: line — the syscall invocation model is implied by the procedure's location in a kernel module. Omit the Invocation: line entirely for these procs.

Prose Conventions

Doc comments are read by people unfamiliar with the change that introduced them. Keep the prose general, accurate, and consistent with the rest of the codebase.

Reuse existing terminology

Use the vocabulary already established in surrounding modules and doc comments. Do not coin new terms or borrow colloquialisms for a concept that already has a name. For example, a value written to a local is "stored" or "saved" (matching loc_storew), not "stashed"; describe what code does plainly rather than labelling it ("load-bearing", "the real check", and similar).

Document the procedure, not the change

Describe the procedure as it currently behaves, for a reader who has never seen the PR that added or modified it. Avoid PR narrative, rationale for a recent fix, and framing such as "this is the X that prevents Y". State what the procedure does and what it guarantees.

Stay at this layer's abstraction

Describe behavior in terms of this procedure and its inputs and outputs. Do not explain how a lower layer (for example the kernel, or a specific syscall) implements or enforces something — that is an implementation detail that may change. Panic bullets in particular state the condition in domain terms ("the asset does not belong to this faucet"), not the mechanism or which layer raises it.

Validation Checklist

  • Description starts with a capitalized present-tense verb and the first sentence ends with a period. Canonical verbs observed in protocol source: Returns, Gets, Computes, Burns, Creates, Increments, Copies, Asserts, Verifies, Hashes, Adds, Removes.
  • Inputs and Outputs use correct stack notation
  • Where section defines every stack item that needs description beyond the Description line, and is omitted entirely when the Description alone covers every item
  • Words are UPPERCASE, felts are lowercase
  • Panics section lists direct asserts and propagated errors
  • Complex panic propagation uses "if fails to verify" shorthand
  • Invocation type specified: exec, call, or dyncall. Kernel (syscall-invoked) procedures are exempt — no Invocation: line.
  • For call and dyncall: padding shown in Inputs/Outputs (see masm-padding skill)
  • Prose reuses existing terminology; no coined terms or colloquialisms
  • Describes the procedure's current behavior, not the change that introduced it
  • No lower-layer (kernel/syscall) implementation details; panic bullets describe conditions in domain terms

Source: 0xMiden/protocol — distributed by TomeVault.

Categories