"Apply deterministic engineering checks to check fixed-width sample arithmetic for signedness, truncation, overflow, and sign-extension errors."
Install
npx skillscat add lucas-belucci-bellini/all-skills-for-the-claude-code/radar-sample-width-signedness-audit Install via the SkillsCat registry.
Radar Sample Width Signedness Audit
Purpose
Check fixed-width sample arithmetic for signedness, truncation, overflow, and sign-extension errors. This workflow turns the concern into a traceable engineering contract for the AERIS-10 project, while keeping simulations, software behavior, and physical measurements clearly distinguished.
Required inputs
Inspect FPGA DSP modules, ADC interfaces, and fixed-point types. Gather the authoritative interface/schema, the selected board or bitstream revision, existing tests, and the expected result. Prefer synthetic input vectors, simulator outputs, host mocks, and de-energized design inspection.
Workflow
- Trace FPGA DSP modules, ADC interfaces, and fixed-point types and record source paths, board/design revision, input/output ownership, and the existing validation entry points.
- Define pass/fail invariants for check fixed-width sample arithmetic for signedness, truncation, overflow, and sign-extension errors.; explicitly document units, widths, timing assumptions, safe defaults, and unsupported states where relevant.
- Exercise Positive/negative full-scale samples, width changes, saturation, wraparound, and cast boundaries with deterministic simulation or mocked hardware. Include normal, boundary, malformed, reset/cancellation, and failure cases without relying on uncontrolled physical targets.
- Produce A width/signedness map with directed test vectors. Run the narrow test and relevant regression suite, capture actual outputs, and separate verified findings from assumptions requiring qualified hardware review.
Validation checklist
- The requirement is represented as an explicit, repeatable assertion or review criterion.
- Boundary values and at least one injected failure are included.
- Units, revision, data origin, and interface timing are stated where they change interpretation.
- Test results cite the exact command/tool run and repository revision.
- Simulation/mock evidence is not misrepresented as hardware measurement or certification.
- RF, power, motion, or external-interface tests stop when the approved safety envelope is unknown or violated.
Required output
Deliver A width/signedness map with directed test vectors. Attach the paths reviewed, fixture or simulation identifiers, test commands and observed results, unresolved safety/compliance questions, and a clear distinction between code-level evidence and physical verification still required.
Boundaries and failure handling
Use synthetic inputs, host tests, simulation, or de-energized inspection wherever possible. Physical RF or powered hardware work requires qualified supervision, an approved lab procedure, regional compliance review, and functioning safety/thermal interlocks. Do not bypass an interlock, increase transmitter power or range as an objective, or turn signal-processing workflows into instructions for targeting people. If the design revision, emissions permission, termination, cooling, or measurement setup is uncertain, stop and document the blocker.
Trigger examples
- Positive: A maintainer needs to check fixed-width sample arithmetic for signedness, truncation, overflow, and sign-extension errors..
- Negative: Ignoring overflow because the nominal test signal remains inside the expected range.