name	workspace-summary
description	Produce a structured inventory of the current workspace — file counts and total size by extension, the largest and most-recently-modified files, and the top-level layout. Use when asked to "summarize", "inventory", "take stock of", "what's in", or "give an overview of" the workspace (works for any folder, not just code repos).

Workspace Summary

Turn the current workspace into a deterministic, structured inventory. Follow these steps in order, then emit the JSON exactly as specified. Keep every command read-only — never create, modify, or delete anything.

Procedure

Enumerate files. Use Glob with pattern **/* from the workspace root. Ignore noise directories: node_modules, .git, target, bin, obj, dist, .venv, __pycache__. Collect the relative path of every regular file.
Group by extension. For each file, take its lowercase extension (or "(none)" when it has none). Count files per extension and, where cheap, sum their sizes. Use Bash (e.g. ls -l, du, wc) only for read-only size/line measurements when needed.
Find the outliers. Identify the 5 largest files (by bytes) and the 5 most-recently-modified files (by mtime). Record path + size for the former, path + an ISO-8601 timestamp for the latter.
Map the top level. List the immediate top-level entries (files and directories) under the workspace root, with a one-line note of what each appears to be (read a README* if present to inform the workspace-level summary).
Emit the inventory. Output exactly one fenced ```json block matching this schema (no extra keys):

{
  "totalFiles": 0,
  "totalDirectories": 0,
  "byExtension": [{ "ext": ".md", "count": 0, "bytes": 0 }],
  "largestFiles": [{ "path": "...", "bytes": 0 }],
  "recentlyModified": [{ "path": "...", "modifiedUtc": "1970-01-01T00:00:00Z" }],
  "topLevel": [{ "name": "...", "kind": "file|dir", "note": "..." }],
  "summary": "one-paragraph overview of what this workspace contains"
}

Notes: counts and the byExtension array must reconcile with totalFiles. Prefer Glob/Read and read-only Bash; if a size is impractical to obtain, set bytes to 0 rather than guessing.

Workspace Summary

Procedure

Enumerate files. Use Glob with pattern **/* from the workspace root. Ignore noise directories: node_modules, .git, target, bin, obj, dist, .venv, __pycache__. Collect the relative path of every regular file.

Group by extension. For each file, take its lowercase extension (or "(none)" when it has none). Count files per extension and, where cheap, sum their sizes. Use Bash (e.g. ls -l, du, wc) only for read-only size/line measurements when needed.

Find the outliers. Identify the 5 largest files (by bytes) and the 5 most-recently-modified files (by mtime). Record path + size for the former, path + an ISO-8601 timestamp for the latter.

Map the top level. List the immediate top-level entries (files and directories) under the workspace root, with a one-line note of what each appears to be (read a README* if present to inform the workspace-level summary).

Emit the inventory. Output exactly one fenced ```json block matching this schema (no extra keys):

{ "totalFiles": 0, "totalDirectories": 0, "byExtension": [{ "ext": ".md", "count": 0, "bytes": 0 }], "largestFiles": [{ "path": "...", "bytes": 0 }], "recentlyModified": [{ "path": "...", "modifiedUtc": "1970-01-01T00:00:00Z" }], "topLevel": [{ "name": "...", "kind": "file|dir", "note": "..." }], "summary": "one-paragraph overview of what this workspace contains" }

Notes: counts and the byExtension array must reconcile with totalFiles. Prefer Glob/Read and read-only Bash; if a size is impractical to obtain, set bytes to 0 rather than guessing.

workspace-summary