Usability pass: tool registry, TUI polish, saved keys and settings

Tools
- Table-driven tool registry with permission classes and default/minimal
  profiles; new tools: edit_file, list_dir, glob, grep, delete_file,
  move_file, todo_write, web_fetch, ask_user (plus read_file paging).
- Built-in Myers diff; write tools return unified diffs.
- Persistent permission rules (.agent/permissions.json), per-turn undo
  snapshots (.agent/undo), AGENTS.md project notes, auto-compact at 80%
  of the model context, cancellable run_shell with process-group kill.

TUI
- Table-driven slash commands: /provider /key /effort /status /tools
  /permissions /yolo /undo /retry /diff /export /copy /rename /delete
  /init /max-steps /todo, with / and @ completion.
- Permission modal with diff and prefix rules, ask_user modal, todo
  panel, header with model/effort/context, bracketed paste, line editing,
  wide-char cursor, NO_COLOR and ASCII fallback; Esc never quits.

Startup and configuration
- API keys are entered in the TUI (hidden) and saved per provider in
  <config>/credentials (0600); env variables are only a fallback.
- Last provider, per-provider model and reasoning effort are remembered
  in <config>/settings.json; --effort/-e and /effort send
  reasoning_effort (low|medium|high|max) to models that accept it.
- DeepSeek preset uses deepseek-flash / deepseek-v4-pro.
- Session flags --new/--continue/--resume/--list-sessions, --yolo,
  --confirm.

Compatibility and tests
- ncurses detection, macOS/Homebrew support, make install, XDG config
  dir, POSIX cleanups; Gitea CI (gcc/clang/asan/musl/macOS).
- Unit tests for diff, fs tools, registry, undo/perms and config;
  e2e fixtures multi_file_refactor and ask_or_todo; vendored cJSON.

fixtures/multi_file_refactor/Makefile

@@ -0,0 +1,20 @@
+CC ?= gcc
+CFLAGS ?= -std=c11 -Wall -Wextra -O2
+
+OBJS = main.o order.o report.o
+
+.PHONY: all clean test
+
+all: shop
+
+shop: $(OBJS)
+	$(CC) $(CFLAGS) -o $@ $(OBJS)
+
+%.o: %.c order.h report.h
+	$(CC) $(CFLAGS) -c -o $@ $<
+
+test: shop
+	./tests/run.sh
+
+clean:
+	rm -f shop $(OBJS)

fixtures/multi_file_refactor/TASK.md

@@ -0,0 +1,15 @@
+# Task: rename across a small C project
+
+`shop` is a tiny C program split over several files. Do the following, then stop:
+
+1. Rename the function `compute_total` to `order_total` everywhere: the
+   declaration, the definition, every call site and the comments. Nothing
+   named `compute_total` may remain anywhere in the project.
+2. `report.c` applies the 10% discount a second time, so the printed total is
+   too low. Fix it so the discount is applied exactly once (`order_total`
+   already applies it).
+3. Run `make test` and make sure it passes.
+4. Summarize the change in two or three sentences.
+
+Find the call sites with the search tools instead of guessing. Edit only
+what the task needs.

fixtures/multi_file_refactor/check.sh

@@ -0,0 +1,11 @@
+#!/usr/bin/env bash
+# Pass iff the rename is complete and the discount bug is fixed.
+set -euo pipefail
+cd "$(dirname "$0")"
+if grep -rn --include='*.c' --include='*.h' compute_total . ; then
+  echo "check: compute_total still present"
+  exit 1
+fi
+grep -q 'int order_total(const int \*prices, int n);' order.h
+make -s clean >/dev/null
+make -s test

fixtures/multi_file_refactor/main.c

@@ -0,0 +1,19 @@
+#include <stdio.h>
+#include <stdlib.h>
+
+#include "order.h"
+#include "report.h"
+
+int main(int argc, char **argv) {
+  int prices[64];
+  int n = 0;
+  for (int i = 1; i < argc && n < 64; i++)
+    prices[n++] = atoi(argv[i]);
+  if (n == 0) {
+    fprintf(stderr, "usage: shop PRICE_CENTS...\n");
+    return 2;
+  }
+  print_report(stdout, prices, n);
+  /* exit status: 0 when compute_total is positive */
+  return compute_total(prices, n) > 0 ? 0 : 1;
+}

fixtures/multi_file_refactor/order.c

@@ -0,0 +1,17 @@
+#include "order.h"
+
+int compute_total(const int *prices, int n) {
+  int sum = 0;
+  for (int i = 0; i < n; i++)
+    sum += prices[i];
+  if (n > 3)
+    sum -= sum / 10; /* bulk discount */
+  return sum;
+}
+
+int count_expensive(const int *prices, int n, int limit) {
+  int c = 0;
+  for (int i = 0; i < n; i++)
+    c += prices[i] > limit;
+  return c;
+}

fixtures/multi_file_refactor/order.h

@@ -0,0 +1,11 @@
+#ifndef ORDER_H
+#define ORDER_H
+
+/* Sum of prices in cents, with the 10% bulk discount applied when there
+ * are more than three items. */
+int compute_total(const int *prices, int n);
+
+/* Number of items that cost more than limit cents. */
+int count_expensive(const int *prices, int n, int limit);
+
+#endif

fixtures/multi_file_refactor/report.c

@@ -0,0 +1,12 @@
+#include "report.h"
+#include "order.h"
+
+void print_report(FILE *out, const int *prices, int n) {
+  /* compute_total gives the discounted sum */
+  int total = compute_total(prices, n);
+  if (n > 3)
+    total -= total / 10;
+  fprintf(out, "items: %d\n", n);
+  fprintf(out, "expensive: %d\n", count_expensive(prices, n, 1000));
+  fprintf(out, "total: %d.%02d\n", total / 100, total % 100);
+}

fixtures/multi_file_refactor/report.h

@@ -0,0 +1,8 @@
+#ifndef REPORT_H
+#define REPORT_H
+
+#include <stdio.h>
+
+void print_report(FILE *out, const int *prices, int n);
+
+#endif

fixtures/multi_file_refactor/tests/run.sh

@@ -0,0 +1,28 @@
+#!/usr/bin/env bash
+# Exit 0 only when shop prints the expected reports.
+set -euo pipefail
+ROOT="$(cd "$(dirname "$0")/.." && pwd)"
+fail=0
+
+expect() {
+  local want="$1"
+  shift
+  local got
+  got="$("$ROOT/shop" "$@")"
+  if [[ "$got" != "$want" ]]; then
+    printf 'FAIL: shop %s\n--- want\n%s\n--- got\n%s\n' "$*" "$want" "$got"
+    fail=1
+  else
+    echo "ok: shop $*"
+  fi
+}
+
+expect $'items: 2\nexpensive: 0\ntotal: 15.00' 500 1000
+expect $'items: 4\nexpensive: 2\ntotal: 36.00' 500 1500 2000 0
+expect $'items: 5\nexpensive: 0\ntotal: 45.00' 1000 1000 1000 1000 1000
+
+if [[ "$fail" -ne 0 ]]; then
+  echo "tests: FAILED"
+  exit 1
+fi
+echo "tests: PASSED"