reviewing-systemverilog

<objective> Review hand-written SystemVerilog (IEEE 1800-2017) through the lens of an experienced FPGA engineer targeting AMD Vivado and Altera Quartus Prime. Produce prioritized findings covering type discipline, always-block usage, naming conventions, and synthesizability. Flag legacy Verilog-2001 patterns as style findings. </objective>

<quick_start> Invoke: /reviewing-systemverilog

Provide the SystemVerilog files you want reviewed (design files, packages, interfaces, testbenches, or any combination). The skill walks through a structured review:

1. Type & Package Review - logic usage, enums, structs, packages, parameterization
2. Always-Block & FSM Review - always_ff/always_comb, blocking vs non-blocking, FSM patterns
3. Structural & Style Review - Port declarations, instantiation, naming conventions
4. Synthesizability Review - Latch detection, width mismatches, tool compatibility
5. Findings Report - Prioritized improvements with idiomatic alternatives

</quick_start>

<essential_principles>

logic everywhere. wire and reg are legacy. logic is the unified type — it works in all contexts. Only use wire for multi-driver nets (tri-state buses).

always_ff and always_comb, never always @. always_ff enforces sequential semantics. always_comb enforces combinational semantics and auto-infers the sensitivity list. always @(posedge clk) and always @(*) are legacy Verilog-2001.

unique case, not bare case. unique case tells the synthesizer all cases are covered and mutually exclusive. priority case when first-match priority encoding is intended. Bare case communicates nothing about designer intent.

Non-blocking (<=) in always_ff, blocking (=) in always_comb. Mixing them up causes simulation/synthesis mismatch. always_ff enforces this; always @(posedge clk) does not.

ANSI-style ports, named connections. Module ports declared in the header, not in a separate body. Instantiations use .port_name(signal), never positional.

Packages for shared types. Enums, structs, typedefs, constants, and functions belong in packages. No `define macros for things that can be parameter or localparam.

</essential_principles>

<intake> What would you like reviewed?

Provide any combination of:

• Design files (.sv, .v) - Modules, always blocks, instantiations
• Package files (.sv) - Type definitions, constants, functions
• Interface files (.sv) - Interface and modport definitions
• Testbench files (.sv) - Stimulus, assertions, coverage
• Constraint files (.xdc, .sdc) - Timing and placement constraints (for context)

You can provide file paths, paste code, or point to a directory.

Wait for the user to provide files before proceeding.

</intake> <routing> After the user provides files, execute the review workflow:

Read ${SKILL_DIR}/references/systemverilog-idioms.md first, then follow ${SKILL_DIR}/workflows/systemverilog-review.md exactly.

User Provides	Reference to Read	Additional Context
Design files	${SKILL_DIR}/references/systemverilog-idioms.md	Full review against all idiom categories
Package files	${SKILL_DIR}/references/systemverilog-idioms.md	Focus on type discipline, naming
Interface files	${SKILL_DIR}/references/systemverilog-idioms.md	Focus on modport, parameterization
Testbench files	${SKILL_DIR}/references/systemverilog-idioms.md	Testbench-specific idioms apply
Mixed	${SKILL_DIR}/references/systemverilog-idioms.md	Review each file in its appropriate mode

</routing>

<reference_index>

File	Purpose
${SKILL_DIR}/references/systemverilog-idioms.md	Comprehensive idiomatic SystemVerilog IEEE 1800-2017 reference
${SKILL_DIR}/workflows/systemverilog-review.md	Step-by-step review procedure with finding format

</reference_index>

<success_criteria> Review is complete when:

• Every file reviewed against idiomatic SystemVerilog standards
• Findings are specific: file, line, what, why, idiomatic alternative
• Findings are prioritized: P0 (incorrect hardware), P1 (quality/maintainability), P2 (style)
• Legacy Verilog-2001 patterns flagged
• Latch and blocking/non-blocking issues identified
• Naming convention deviations noted
• Summary table of all findings delivered

</success_criteria>