alexbeatnik

Add a new verb/command to the ManulEngine (Go) .hunt DSL. Use when extending the parser with a new CommandType, wiring runtime dispatch, and writing parser + runtime tests. Covers the full multi-file change checklist.

Add or modify low-level browser capabilities in ManulEngine (Go). Use when implementing a new Page interface method, touching Chrome lifecycle, CDP commands, input dispatch, or JS evaluation. Covers the interface-to-backend wiring pattern.

Enforce ManulEngine (Go)'s concurrency contract when editing runtime/, cdp/, worker/, or anything that spawns goroutines. Use when adding shared state, introducing a goroutine, modifying the Worker/Pool API, or touching CDP transport internals.

Add or modify runtime configuration fields in ManulEngine (Go). Use when introducing a new tunable engine parameter, wiring CLI flags, or env-var overrides. Covers the 3-layer priority chain and pointer-based JSON overlay.

Debug or extend the ManulEngine (Go) DSL preprocessor. Use when working with @import resolution, USE/CALL block expansion, variable interpolation, or scoped variable behavior. Covers the 5-level precedence hierarchy and circular-import detection.

Add or debug Go-level extensions — RegisterCustomControl and RegisterGoCall. Use when wiring a custom element handler, a CALL GO target, or when writing tests that involve the extension registry.

Author or review ManulEngine (Go) .hunt DSL files. Use when the user asks to write, modify, or review a .hunt file, or when generating automation for a web flow that will be run by the `manul` CLI.

Add metrics to execution reports or debug the explainability pipeline. Use when modifying pkg/explain data types, updating HTML report generation, or wiring new scorer signals into the report output. Covers per-hunt reports and aggregate index.html.

Guidelines for modifying the VS Code Chat participant (@manulai) in src/copilotChatParticipant.ts. Covers streaming, slash commands, auto-approve state, agent mode awareness, history handling, and keeping it independent from legacy code.

Guidelines for modifying the workspace agent instructions reader in src/agentInstructionsReader.ts. Covers file discovery, AGENTS.md/CLAUDE.md parsing, and injection into the Copilot Chat system prompt.

Guidelines for versioning, packaging, and releasing the ManulAI VS Code extension. Covers version bumps, VSIX builds, documentation sync, and release checks.

Guidelines for modifying Ollama streaming and reasoning extraction in src/ollamaStreamParser.ts. Covers NDJSON parsing, <think> tag extraction, and integration with the Copilot Chat participant.

Guidelines for modifying the ManulAI Settings webview panel in src/settingsPanel.ts. Covers Activity Bar registration, model fetching from Ollama, webview HTML, message passing, and VS Code settings updates.

Guidelines for modifying the workspace skills reader in src/skillsReader.ts. Covers skill discovery, frontmatter parsing, and injection into the Copilot Chat system prompt.

Atomically bump the ManulAI extension version across all 6 required locations (package.json, package-lock.json ×2 occurrences, README.md What's New, README-dev.md Release Notes, CLAUDE.md, .github/copilot-instructions.md) and verify the build still compiles. Enforces the CLAUDE.md rule that packaging version changes update every one of these files in the same change.

Run scripts/debug-agent.mjs against the full validated Ollama baseline (phi4-mini:3.8b, llama3.1:8b, gemma4:latest, gemma4:31b, qwen3-coder:30b) for a single user prompt. Collects pass/fail per model and reports JSONL log paths. Use this for regression testing after any change to the agent loop, recovery/nudge logic, mandate, or tool handlers.

Add a new configuration setting to ManulEngine end-to-end — EngineConfig field, MANUL_* env var, JSON config key, prompts.py global, and the MANUL_CONFIG_CONTRACT.md entry — honoring the precedence layering. Invoke when the user says "add a config option", "new setting", "new MANUL_ env var", "make X configurable", "додай налаштування", "новий конфіг". A new public ctor kwarg is "Ask first" per CLAUDE.md.

Author a @custom_control handler that overrides ManulEngine's element resolution for a specific (page, target) pair — for widgets the heuristics can't drive (React date pickers, canvas, custom web components). Wires the handler into a controls/ module, aligns the page label with pages/*.json, and adds a test. Invoke when the user says "add a custom control", "override resolution for X", "the heuristics can't click this widget", "handle this React/canvas component", "додай кастомний контрол".

Add a new DSL step kind (e.g. WAIT FOR DOWNLOAD, EXTRACT TABLE, …) to ManulEngine end-to-end. Wires classify_step + dispatch in BOTH the main run_mission loop AND the conditional-branch _dispatch_step executor, then adds a synthetic-DOM test. Invoke when the user says "add step", "new DSL keyword", "support a new action".

Run the ManulEngine synthetic-DOM test suite (manul_engine/test/test_*.py — 54 function-based suites, NOT unittest) and surface failures concisely. Invoke when the user says "run tests", "run the suite", "are tests passing", "запусти тести", or after non-trivial code changes to core.py / scoring.py / helpers.py / actions.py / cli.py before reporting work as done.

Create a working Python virtualenv for ManulEngine and run its tests. ManulEngine drives a system Chrome/Chromium directly over CDP — there is no Playwright and no bundled-browser download. Invoke when the user says "create venv", "setup env", "set up the environment", "install deps", "створи венв", "налаштуй середовище", or when a test run fails with "No module named 'websockets'" or "Chrome/Chromium not found".

Verify and reconcile the frozen public-surface contracts under contracts/MANUL_*_CONTRACT.md against the current code. Invoke when the user says "check contracts", "are contracts up to date", "sync contracts", or after any change that touches the DSL grammar, CLI flags, ManulEngine/ManulSession ctor, EngineConfig fields, hooks, reporter JSON, or scoring weights.

Bump the ManulEngine version atomically across all 17 files using bump_version.py (the single source of truth is pyproject.toml). Invoke when the user says "bump to X.Y.Z", "онови версію", "cut a release", or any phrasing that implies a version change. Refuses to run if the working tree is on an inconsistent version base.

Version bumps and releases of ClamAV UI — AssemblyInfo.cs, the release.yml workflow, vX.Y.Z branch flow, and the self-update constraints a release must respect. Use when bumping the version, preparing a release, or changing anything about update/release plumbing.

Retake the four README screenshots (dashboard, logs, quarantine, settings) by driving the running ClamAV UI app through UI Automation. Use when asked for new/updated README screenshots or to capture the app's pages.

Launch and manually verify ClamAV UI after a change — single-instance/tray behavior, what to click through, and the Windows Defender/EICAR gotcha. Use after building when a change needs to be seen working in the real app.

Add or change user-visible text in ClamAV UI — Lang.cs keys, English+Ukrainian translations, ApplyLanguage re-texting. Use whenever any UI label, button, log line, tray balloon, status message, or dialog string is added or edited.

Add a persisted option to ClamAV UI — a settings.ini key with LoadSettings/SaveSettings wiring, migration for old settings files, and a control on the Settings page (absolute-positioned card layout). Use when adding a toggle, selector, or any state that must survive restarts.

Add or run unit tests for ClamAV UI — the zero-dependency runner in tests/TestFramework.cs, making logic testable via internal static helpers, TempDir usage. Use when writing tests, making code testable, or investigating test failures.

Add support for a new audio file extension (e.g. .aiff, .wma) to AudioToText, wiring the file-picker filter and deciding native-vs-ffmpeg-conversion. Use when the user wants the app to accept another audio format.

Build, vet, format-check and static-analyze the AudioToText Go project. Use after any code change to confirm it compiles and is clean before finishing or committing.

Retake the four README screenshots (dashboard, logs, quarantine, settings) by driving the running AV app through UI Automation. Use when asked for new/updated README screenshots or to capture the app's pages.