agentleFS
iktok90-designiktok90-design/ai-smart-contract-auditor/skills/inline-assembly/SKILL.md

inline-assembly

Skill36 starsChanged 5 months agoMITView on GitHub

Detect bugs in inline Yul / assembly — manual memory mismanagement, free-memory-pointer corruption, return-data manipulation, missing return-data-size checks, dirty-bits in narrow types. Activate on any `assembly { … }` block, Yul code, Solady-style assembly usage.

The file

We split this skill into the rules it gives an agent

7 rules. About 0 are common, 3 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    After writing memory, update mload(0x40) to past the write.

    line 26
  2. Very specific:

    Check returndatasize() before copying.

    line 39
  3. Distinctive:

    Prefer audited primitives (OZ, Solady) — don't roll your own assembly.

    line 80
  4. Distinctive:

    Document every assembly block: invariants, memory map, why.

    line 81
  5. Very specific:

    Always update mload(0x40) after writes.

    line 82
  6. Very specific:

    Check returndatasize() before returndatacopy.

    line 83
  7. Distinctive:

    Test with edge-case inputs (zero-length return, oversize return, dirty high bits).

    line 85

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Inline assembly / Yul auditor

When this applies

  • Any assembly { … } block
  • Pure Yul contracts
  • Solady's heavy use of optimized assembly
  • Custom delegatecall wrappers, manual memory copy, custom signature verifiers
  • Gas-optimization-driven assembly substitutions of Solidity primitives

Detection patterns

Free-memory-pointer not updated after allocation (HIGH)

assembly {
    let ptr := mload(0x40)
    mstore(ptr, value)
    // ← didn't bump 0x40, next allocation corrupts
}

After writing memory, update mload(0x40) to past the write.

Memory clobbering via reused scratch space (HIGH)

Yul scratch is 0x00-0x3F. External calls / Solidity assignments may overwrite. Don't store across foreign-code boundaries.

Return data not size-checked (HIGH)

assembly {
    let ok := call(gas(), target, 0, in, insz, 0, 32)
    returndatacopy(0, 0, 32)
    let r := mload(0)   // ← if target returned < 32 bytes, r has trailing memory garbage
}

Check returndatasize() before copying.

Dirty high bits in narrow types (HIGH)

A uint8 read from calldata via calldataload has the high 248 bits unmasked. Mask with and(..., 0xff) before comparing.

Returning attacker-controlled memory (HIGH)

assembly { return(0, calldatasize()) }   // ← returns calldata; if function spec says bytes32, parsers blow up

Mishandling 0x40 / 0x60 (FMP and zero slot) (HIGH)

Writing to 0x60 (zero slot) is a known footgun — that slot must stay zero. Solidity uses it in mappings.

mstore to slot 0/0x20 not cleared after a call (MEDIUM-HIGH)

Reused scratch leaks data across functions.

Loop bounds in assembly without checks (HIGH)

Easy to write infinite loops or array-bound-violators.

Wrong opcode (callcode, suicide, gas costs) (HIGH)

Legacy opcodes still parseable but deprecated. callcode is the wrong delegatecall; suicide is selfdestruct.

Address pointers larger than 20 bytes (HIGH)

address masked at 20 bytes; assembly users sometimes treat as full 32 bytes and leave dirty bits.

Severity rubric

PatternSeverity
Free-memory-pointer corruptionHigh
Unchecked returndatasizeHigh
Dirty high bits on narrow typeHigh
Wrong/legacy opcodeHigh
Writes to 0x60 (zero slot)High
Reused scratch across foreign callHigh
Assembly returning misformatted ABI dataHigh
Pure-Yul math with no overflow proofHigh
Comment-less assembly block in production codeLow (maintainability)

Remediation patterns

  • Prefer audited primitives (OZ, Solady) — don't roll your own assembly.
  • Document every assembly block: invariants, memory map, why.
  • Always update mload(0x40) after writes; explicitly zero scratch after use.
  • Check returndatasize() before returndatacopy.
  • Mask narrow types: and(x, 0xff) for uint8, and(x, 0xffff) for uint16, etc.
  • Test with edge-case inputs (zero-length return, oversize return, dirty high bits).

False-positive notes

  • Single-line read of fixed slot (e.g. chainid(), gasprice(), caller()) is safe.
  • Solady's published assembly libraries are audited; using them as-is is fine — verify version pinning.

Related

  • [[assembly-auditor]] (subagent)
  • [[unchecked-calls]]
  • [[storage-layout]]

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.