Use this agent to diagnose and fix a Delphi crash report (.mad file) end-to-end up to and INCLUDING a successful build, for any product registered in the /light-bug-MadShi pipeline (BioniX, ...). Stops before release. Reads the .mad, parses the madExcept call stack, scans the product's Fixes Log for duplicates…
Audit a Delphi project's own classes (INCLUDING forms) for bad shapes — god classes, anemic data classes, misplaced responsibility, fat config records, shared mutable singletons, Ex-downcast holders. Reports ranked reshapes (extract-class, move-method, replace-singleton); never refactors. Reads whole classes in its…
Audit a Delphi project's own units (files) for SHALLOW modules that should merge into one deep module — information leakage, temporal decomposition, pass-through layers, always-paired usage. Reports ranked merge candidates; never merges. Reads whole units in its own context and returns a ranked report, keeping the…
Use this agent to audit Delphi FMX source files for cross-platform compatibility issues. Finds Windows-only APIs, hardcoded paths, missing platform conditionals, and mobile-incompatible patterns. Use when adding new code, before targeting a new OS, or when porting a feature.
Use this agent when you need to scan imported or 3rd-party Delphi code for style compliance, common mistakes, and dangerous patterns. Do NOT use it for our own project code — use light-review-Full instead.
Use this agent when you need to compile a Delphi project and verify that code changes build successfully. This includes after making code modifications, refactoring, adding new units, or any time you need to check that the codebase is in a compilable state.
Use this agent to fix Delphi vocabulary and writing-clarity issues in Markdown documentation files. This agent is the workhorse behind the /light-md-DelphiIdiom skill — it reads target MD files whole, applies vocabulary swaps from the dictionary, and applies bounded clarity rewrites with a high confidence bar. Default…
Scan a Delphi project's markdown docs (CLAUDE.md, README.md, docs/.md) for drift against the current code — stale class names, file paths, settings keys, procedure signatures, architecture claims — then apply surgical Edit fixes. Reads whole docs and verifies each claim against the code in its own context, returning a…
Use this agent to prune and tighten a CLAUDE.md (or any agent/skill instruction markdown): cut bloat and duplication, delete stale/wrong/contradictory rules, relocate misplaced content to the right layer (agent/skill/hook/scoped file), and reorder for adherence — without losing load-bearing information. It verifies…
Audit a Delphi DUnitX or classic-DUnit test suite for FAKE or WEAK tests — tests that pass without verifying the behavior they name (zero assertions, Assert.Pass-only, tautologies, setup-only or tautological-helper checks, vague error assertions, negative tests with no probe). Reads every [Test] or published method…
Stage 1 of the Delphi review pipeline — a thorough, critical code review of Delphi source files. This is NOT a style checker — it reads code to understand intent, then verifies correctness. Use it for our own project code when you need a real review. Do NOT use it for 3rd-party imports (use light-code-StyleChecker for…
Stage 2 of the Delphi review pipeline — counter-analyze the findings of a prior light-review-step1 run. Drops false positives (verifies every claim by reading the actual file, the declaration of every named type, and at least one caller — plus Embarcadero docs for RTL/FMX/VCL APIs), then applies the fixes that survive…
Stage 3 of the Delphi review pipeline — verify code fixes hold up, revert the ones that don't, then compile. For each edit it confirms the fix matches the stated reasoning, didn't change observable behavior, didn't miss call sites, and didn't break DFM/FMX bindings; then it runs the project's tests (large change) or…
Use this agent when the user has written or modified HTML, CSS, or JavaScript code and needs it reviewed for errors, formatting issues, best practices violations, accessibility problems, or cross-browser compatibility concerns. This agent should be launched after writing or editing web frontend code.
Use this agent to summarize a YouTube transcript text file. The agent first cleans the transcript via the PowerShell script (if not already cleaned), reads the cleaned text, and produces a structured summary. If the transcript is about programming, the summary includes a section adapting the ideas to the user's Delphi…
★not rated 16 yesterdayA87 tokens
originalMPL-2.0
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: