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/ask_or_todo/TASK.md

@@ -0,0 +1,20 @@
+# Task: three small scripts, planned with a task list
+
+Create these files in the workspace root:
+
+1. `greet.sh` — prints `Hello, NAME!` where NAME is the first argument, or
+   `Hello, world!` when no argument is given.
+2. `count.sh` — prints the number of lines in the file given as the first
+   argument (just the number). Exits with status 2 and prints a usage line to
+   stderr when no argument is given.
+3. `SCRIPTS.md` — one short section per script explaining how to run it.
+
+Both scripts must be executable POSIX shell scripts (`#!/bin/sh`).
+
+Rules:
+- Start by calling `todo_write` with one item per file, then keep the list
+  up to date (mark each item done as you finish it). Every item must be done
+  at the end.
+- Nobody is available to answer questions in this run. If `ask_user` fails,
+  pick a sensible default and continue.
+- Test both scripts with `run_shell` before you finish.

fixtures/ask_or_todo/check.sh

@@ -0,0 +1,34 @@
+#!/usr/bin/env bash
+# Pass iff the scripts work and the agent kept a finished task list.
+set -euo pipefail
+cd "$(dirname "$0")"
+fail() { echo "check: $*"; exit 1; }
+[[ -x greet.sh && -x count.sh ]] || fail "scripts missing or not executable"
+[[ "$(./greet.sh)" == "Hello, world!" ]] || fail "greet.sh without argument"
+[[ "$(./greet.sh Ada)" == "Hello, Ada!" ]] || fail "greet.sh Ada"
+[[ "$(./count.sh notes.txt | tr -d ' ')" == "3" ]] || fail "count.sh notes.txt"
+set +e
+./count.sh >/dev/null 2>&1
+rc=$?
+set -e
+[[ "$rc" -eq 2 ]] || fail "count.sh without argument should exit 2 (got $rc)"
+[[ -s SCRIPTS.md ]] || fail "SCRIPTS.md missing"
+grep -q greet.sh SCRIPTS.md && grep -q count.sh SCRIPTS.md ||
+  fail "SCRIPTS.md does not mention both scripts"
+# The session must show todo_write ending with every item done.
+python3 - <<'PY' || fail "no finished todo_write list in the session"
+import glob, json, re, sys
+last = None
+for path in glob.glob(".agent/*.jsonl") + glob.glob(".agent/sessions/*.jsonl"):
+    for line in open(path, encoding="utf-8"):
+        try:
+            m = json.loads(line)
+        except ValueError:
+            continue
+        c = m.get("content") or ""
+        r = re.search(r"todo list updated: (\d+)/(\d+) done", c)
+        if r:
+            last = (int(r.group(1)), int(r.group(2)))
+sys.exit(0 if last and last[1] >= 3 and last[0] == last[1] else 1)
+PY
+echo "check: PASSED"

fixtures/ask_or_todo/notes.txt

@@ -0,0 +1,3 @@
+first
+second
+third

fixtures/big_file/Makefile

@@ -0,0 +1,5 @@
+.PHONY: check
+
+check:
+	@test -f lines.txt || { echo "lines.txt missing"; exit 1; }
+	@seq -f 'line %04g' 1 500 | cmp - lines.txt && echo "check: PASSED ($$(wc -c < lines.txt) bytes)"

fixtures/big_file/TASK.md

@@ -0,0 +1,8 @@
+# 任务：写一个大于 4KB 的文件
+
+请用 `write_file` 工具（不要用 shell 的 seq/printf/循环生成）一次性写出 `lines.txt`：
+
+- 共 500 行，第 i 行是 `line ` 加上 4 位补零的 i，例如第 1 行 `line 0001`，第 500 行 `line 0500`；
+- 每行以换行结尾，文件总长 5000 字节。
+
+然后运行 `make check` 验证；若失败，读输出修正后再验证。最后用一句话说明结果。

fixtures/big_file/check.sh

@@ -0,0 +1,4 @@
+#!/usr/bin/env bash
+set -euo pipefail
+cd "$(dirname "$0")"
+make check

fixtures/makefile_tab/TASK.md

@@ -0,0 +1,11 @@
+# 任务：为 hello 项目写 Makefile
+
+工作区里只有 `hello.c` 和 `tests/expect.txt`，还没有 Makefile。请：
+
+1. 新建 `Makefile`（注意：命令行必须用 TAB 缩进），包含以下目标：
+   - `all`（默认目标）：用 `cc -std=c11 -Wall -Wextra -O2 -o hello hello.c` 编译出 `hello`；
+   - `test`：依赖 `hello`，运行 `./hello > tests/actual.txt`，再用 `diff -u tests/expect.txt tests/actual.txt` 比较；
+   - `clean`：删除 `hello` 和 `tests/actual.txt`；
+   - 把 `all test clean` 声明为 `.PHONY`。
+2. 运行 `make clean && make test`，确认通过。
+3. 用一句话说明结果。

fixtures/makefile_tab/check.sh

@@ -0,0 +1,10 @@
+#!/usr/bin/env bash
+# Pass iff the agent wrote a working Makefile (recipes need real TABs).
+set -euo pipefail
+cd "$(dirname "$0")"
+test -f Makefile
+grep -q "$(printf '^\t')" Makefile
+make clean >/dev/null
+make test
+make clean >/dev/null
+test ! -e hello

fixtures/makefile_tab/hello.c

@@ -0,0 +1,7 @@
+#include <stdio.h>
+
+int main(void) {
+  printf("hello, cst\n");
+  printf("tab\there\n");
+  return 0;
+}

fixtures/makefile_tab/tests/expect.txt

@@ -0,0 +1,2 @@
+hello, cst
+tab	here

fixtures/mini_calc/check.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+# Pass iff the task is done (used by scripts/e2e_fixture.sh).
+set -euo pipefail
+cd "$(dirname "$0")"
+make test

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"

fixtures/snake_cpp/check.sh

@@ -0,0 +1,5 @@
+#!/usr/bin/env bash
+set -euo pipefail
+cd "$(dirname "$0")"
+make snake
+test -x snake

fixtures/test_fix_test/Makefile

@@ -0,0 +1,14 @@
+CC ?= cc
+CFLAGS ?= -std=c11 -Wall -Wextra -O2
+
+.PHONY: test clean
+
+test_strutil: strutil.c strutil.h tests/test_strutil.c
+	$(CC) $(CFLAGS) -o $@ strutil.c tests/test_strutil.c
+
+test: test_strutil
+	@echo run >> .test_runs
+	./test_strutil
+
+clean:
+	rm -f test_strutil

fixtures/test_fix_test/TASK.md

@@ -0,0 +1,6 @@
+# 任务：先跑测试，再修，再跑测试
+
+1. 先运行 `make test`，看哪些用例失败（此时不要改代码）。
+2. 阅读 `strutil.c`，修复导致失败的 bug（不要修改 `tests/` 下的文件）。
+3. 再次运行 `make test`，确认全部通过。
+4. 用简短中文说明你修了什么。

fixtures/test_fix_test/check.sh

@@ -0,0 +1,12 @@
+#!/usr/bin/env bash
+# Pass iff tests pass now AND the agent really ran `make test` at least twice
+# (before and after the fix; an over-eager duplicate filter would skip one).
+set -euo pipefail
+cd "$(dirname "$0")"
+runs=$(wc -l < .test_runs 2>/dev/null || echo 0)
+echo "agent test runs: $runs"
+git_clean_tests=1
+grep -q 'check_vowels("Queue", 4);' tests/test_strutil.c || git_clean_tests=0
+[ "$git_clean_tests" = 1 ] || { echo "tests were modified"; exit 1; }
+[ "$runs" -ge 2 ] || { echo "make test ran fewer than 2 times"; exit 1; }
+make test

fixtures/test_fix_test/strutil.c

@@ -0,0 +1,23 @@
+#include "strutil.h"
+
+#include <ctype.h>
+#include <string.h>
+
+void str_reverse(char *s) {
+  size_t n = strlen(s);
+  for (size_t i = 0; i < n / 2; i++) {
+    char t = s[i];
+    s[i] = s[n - i];
+    s[n - i] = t;
+  }
+}
+
+int count_vowels(const char *s) {
+  int k = 0;
+  for (; *s; s++) {
+    char c = (char)tolower((unsigned char)*s);
+    if (c == 'a' || c == 'e' || c == 'i' || c == 'o')
+      k++;
+  }
+  return k;
+}

fixtures/test_fix_test/strutil.h

@@ -0,0 +1,7 @@
+#ifndef STRUTIL_H
+#define STRUTIL_H
+
+void str_reverse(char *s);
+int count_vowels(const char *s);
+
+#endif

fixtures/test_fix_test/tests/test_strutil.c

@@ -0,0 +1,39 @@
+#include "../strutil.h"
+
+#include <stdio.h>
+#include <string.h>
+
+static int fails;
+
+static void check_rev(const char *in, const char *want) {
+  char buf[64];
+  snprintf(buf, sizeof(buf), "%s", in);
+  str_reverse(buf);
+  if (strcmp(buf, want) != 0) {
+    printf("FAIL str_reverse(\"%s\") = \"%s\", want \"%s\"\n", in, buf, want);
+    fails++;
+  }
+}
+
+static void check_vowels(const char *in, int want) {
+  int got = count_vowels(in);
+  if (got != want) {
+    printf("FAIL count_vowels(\"%s\") = %d, want %d\n", in, got, want);
+    fails++;
+  }
+}
+
+int main(void) {
+  check_rev("abc", "cba");
+  check_rev("ab", "ba");
+  check_rev("", "");
+  check_vowels("education", 5);
+  check_vowels("RHYTHM", 0);
+  check_vowels("Queue", 4);
+  if (fails) {
+    printf("%d test(s) failed\n", fails);
+    return 1;
+  }
+  printf("all tests passed\n");
+  return 0;
+}