TypeScript and JavaScript agent workflows with TGlider

Practical examples for using TGlider with a coding agent. Each one asks the agent to gather structured TS/JS facts before editing.

Map a feature across a monorepo

Ask

> Load this workspace, find the entry points for the checkout feature, then map the packages, exports, components, services, tests, and callers involved before proposing edits.

Why it helps

JS/TS features often cross package boundaries and barrel exports. TGlider helps the agent map ownership before reading random files.

Tools: load, get_workspace_root, find_code, list_exports, find_callers

Check package import usage

Ask

> Find every project and file importing zod, then summarize whether the usage is runtime validation, type inference, or test-only support.

Why it helps

Package usage can be scattered across apps and packages. Semantic package importer queries provide concrete import evidence quickly.

Tools: semantic_query, get_package_info, get_file_contents

Inspect public API surface before changing exports

Ask

> List the exported API surface for the ui package, then find which exports are referenced by the web app before changing the public component name.

Why it helps

Changing an exported symbol can break downstream packages even when local tests pass. API and reference tools make that impact visible.

Tools: list_project_apis, list_exports, semantic_query, analyze_change_impact

Investigate diagnostics without reading build logs

Ask

> Run diagnostics, group hotspots by file, inspect the top failing source ranges, and propose the smallest fix plan.

Why it helps

TypeScript output can be noisy in monorepos. TGlider returns compact diagnostics and grouped hotspots for agent triage.

Tools: get_diagnostics, diagnostic_hotspots, get_file_contents

Find implementations and callers

Ask

> Find the implementations of PaymentGateway, then trace callers of the selected implementation before changing retry behavior.

Why it helps

Interfaces, classes, and aliases can hide behavior behind similarly named code. Semantic implementation and call graph tools reduce false matches.

Tools: find_code, find_implementations, find_callers, get_outgoing_calls

Preview a rename safely

Ask

> Resolve useCustomerStore, preview renaming it to useAccountStore, summarize affected files and diagnostics risk, then apply only if the plan is clean.

Why it helps

Language-service rename avoids changing comments, strings, or unrelated symbols that share the same text.

Tools: find_code, compute_rename_edits, analyze_change_impact, rename_symbol

Clean unused private code cautiously

Ask

> Find likely-unused private symbols and parameters in the app project, then summarize safe cleanup candidates without touching public exports.

Why it helps

Absence of text matches is weak evidence. Conservative unused-code tools report candidates with symbol context and exclusions.

Tools: find_unused_symbols, find_unused_parameters, get_symbol_info

Keep semantic reads aligned with unsaved editor state

Ask

> Open this modified file as a live document, run structure and diagnostics against the in-memory text, then close it when done.

Why it helps

Agents often reason about edits before they are saved. Live-document tools prevent stale disk reads from mixing with current editor text.

Tools: open_document, get_structure, get_diagnostics, close_document

↑/↓ NavigateEnter OpenSpace Expand