planning: write v0.4 sprint doc, promote NUG-015 to backlog

planning/README.md

@@ -15,7 +15,8 @@ under five minutes.
 | [audit.md](audit.md) | Snapshot of the codebase: what's built, what's tested, what's drifted from docs, what's stubbed |
 | [roadmap.md](roadmap.md) | Strategic shaping — themes, v0.3 / v0.4 release framing, design-pending items |
 | [backlog.md](backlog.md) | Prioritised tickets (NUG-001…NUG-014) with acceptance criteria, effort, P-priority, labels |
-| [next-sprint.md](next-sprint.md) | The 4 highest-priority tickets (NUG-001, 002, 003, 010), fully scoped for immediate pickup — the v0.3 sprint plan |
+| [next-sprint.md](next-sprint.md) | The 4 highest-priority tickets (NUG-008, 011, 009, 004), fully scoped for immediate pickup — the v0.4 sprint plan |
+| [sprint-v0.3.md](sprint-v0.3.md) | Archived v0.3 sprint plan (NUG-001, 002, 003, 010) |
 | [issue-1.md](issue-1.md) | Standalone issue: `filebrowser` `cat` op only takes a single path — model must fan out to N tool calls to read N files, which burns the 16-iteration loop budget and looks like flakiness |
 
 ## How the IDs map to ROADMAP.md
@@ -37,6 +38,8 @@ under five minutes.
 | NUG-013 | #17 | Bench: flakiness report |
 | NUG-014 | #15 | Semantic search over memory + sessions |
 
+| NUG-015 | — | `filebrowser cat` multi-path support |
+
 Items #4 (full hooks framework), #6 (MCP), #12–14 (agent configs / skills /
 subagents) are not ticketed yet — they need a design-doc round before they
 can be sliced into actionable work. See [roadmap.md](roadmap.md) for the

planning/backlog.md

@@ -228,6 +228,37 @@ Vector index over `memory.db` rows and session message bodies. Surfaced as a new
 
 ---
 
+## Tool ergonomics (parallel track)
+
+### NUG-015 · `filebrowser cat` multi-path support · P2 · S · feature
+**Roadmap #:** —
+**Promoted from:** `planning/issue-1.md`
+
+Today `filebrowser`'s `cat` op takes a single `path: string`. Reading N files
+requires N separate tool calls, burning the 16-iteration loop budget and
+producing apparent flakiness when the model drops to partial reads.
+
+**Acceptance criteria:**
+- `cat` with `path: "a.txt"` returns the existing `{path, content, size}`
+  shape (back-compat).
+- `cat` with `path: ["a.txt", "b.txt"]` returns
+  `{results: {"a.txt": {content, size}, "b.txt": {content, size}}}` with
+  per-file errors slotted under their key as `{"error": "..."}`.
+- Schema's `path` field description updated to document both forms.
+- Total response capped at a configurable byte budget (default 256 KiB)
+  with truncation marked per-file; prevents one large file from blowing
+  the context window.
+- Tests in `tests/tools/test_filebrowser.py` covering: single-string
+  back-compat, list of two existing files, list with one missing file,
+  list exceeding byte budget.
+
+**Files likely touched:** `src/nugget/tools/filebrowser.py`,
+`tests/tools/test_filebrowser.py`.
+
+**Out of scope:** glob expansion (`*.py`), streaming or chunked reads.
+
+---
+
 ## Bench enhancements (parallel track)
 
 ### NUG-012 · Bench: prompt-variant sweeping · P3 · M · feature