@team-agent/installer 0.2.11 → 0.3.1

package/crates/team-agent/src/model/testdata/team.spec.golden.yaml

@@ -0,0 +1,206 @@
+version: 1
+team:
+  name: "teamspec-full-example"
+  mode: "supervisor_worker"
+  objective: "Build, research, review, and document a code change with Codex CLI workers."
+  workspace: "."
+leader:
+  id: "leader"
+  role: "leader"
+  provider: "codex"
+  model: null
+  tools:
+    - "fs_read"
+    - "fs_list"
+    - "mcp_team"
+    - "provider_builtin"
+  context_policy:
+    keep_user_thread: true
+    receive_worker_outputs: "structured_only"
+    max_worker_result_tokens: 4000
+agents:
+  - id: "codex_implementer"
+    role: "implementation_engineer"
+    provider: "codex"
+    model: null
+    working_directory: "."
+    system_prompt:
+      inline: |
+        You are the implementation worker. Make focused code changes, run relevant tests,
+        and report a result_envelope_v1 with changed files, tests, risks, artifacts, and next actions.
+      file: null
+    tools:
+      - "fs_read"
+      - "fs_write"
+      - "fs_list"
+      - "execute_bash"
+      - "git_diff"
+      - "mcp_team"
+      - "provider_builtin"
+    permission_mode: "restricted"
+    preferred_for:
+      - "implementation"
+      - "bug_fix"
+      - "test"
+    avoid_for:
+      - "final_risk_signoff"
+    output_contract:
+      format: "result_envelope_v1"
+      required_fields:
+        - "task_id"
+        - "status"
+        - "summary"
+        - "artifacts"
+  - id: "codex_researcher"
+    role: "researcher"
+    provider: "codex"
+    model: null
+    working_directory: "."
+    system_prompt:
+      inline: |
+        You are the research worker. Prefer read-only analysis and summarize findings
+        as result_envelope_v1. Do not edit files.
+      file: null
+    tools:
+      - "fs_read"
+      - "fs_list"
+      - "network"
+      - "mcp_team"
+      - "provider_builtin"
+    permission_mode: "restricted"
+    preferred_for:
+      - "research"
+      - "architecture"
+      - "docs"
+    avoid_for:
+      - "implementation"
+    output_contract:
+      format: "result_envelope_v1"
+      required_fields:
+        - "task_id"
+        - "status"
+        - "summary"
+        - "artifacts"
+  - id: "codex_reviewer"
+    role: "code_reviewer"
+    provider: "codex"
+    model: null
+    working_directory: "."
+    system_prompt:
+      inline: |
+        You are the reviewer. Find correctness, regression, security, and missing-test risks.
+        Stay read-only unless the leader explicitly changes your permissions.
+      file: null
+    tools:
+      - "fs_read"
+      - "fs_list"
+      - "git_diff"
+      - "mcp_team"
+      - "provider_builtin"
+    permission_mode: "restricted"
+    preferred_for:
+      - "review"
+      - "risk_check"
+    avoid_for:
+      - "implementation"
+    output_contract:
+      format: "result_envelope_v1"
+      required_fields:
+        - "task_id"
+        - "status"
+        - "summary"
+        - "artifacts"
+routing:
+  default_assignee: "leader"
+  rules:
+    - id: "implementation-to-codex"
+      when: "task.type in [\"implementation\", \"bug_fix\", \"test\"]"
+      assign_to: "codex_implementer"
+      priority: 100
+    - id: "research-to-codex"
+      when: "task.type in [\"research\", \"architecture\", \"docs\"]"
+      assign_to: "codex_researcher"
+      priority: 90
+    - id: "review-to-codex"
+      when: "task.type in [\"review\", \"risk_check\"]"
+      assign_to: "codex_reviewer"
+      priority: 90
+communication:
+  protocol: "mcp_inbox"
+  topology: "leader_centered"
+  worker_to_worker: false
+  ack_timeout_sec: 60
+  result_format: "result_envelope_v1"
+  message_store:
+    sqlite: ".team/runtime/team.db"
+    mirror_files: ".team/messages"
+runtime:
+  backend: "tmux"
+  display_backend: "none"
+  session_name: "teamspec-full-example"
+  auto_launch: true
+  require_user_approval_before_launch: true
+  dangerous_auto_approve: false
+  max_active_agents: 3
+  startup_order:
+    - "codex_implementer"
+    - "codex_researcher"
+    - "codex_reviewer"
+context:
+  state_file: "team_state.md"
+  artifact_dir: ".team/artifacts"
+  log_dir: ".team/logs"
+  summarization:
+    worker_full_logs: "retain_outside_leader_context"
+    state_update: "after_each_result"
+tasks:
+  - id: "task_research"
+    title: "Read the task context and identify design risks."
+    type: "research"
+    assignee: null
+    deps: []
+    acceptance:
+      - "Result envelope includes summary and risks."
+    status: "pending"
+    requires_tools:
+      - "fs_read"
+    files:
+      - "**/*"
+    risk: "medium"
+    retry_limit: 1
+    human_confirmation: false
+  - id: "task_impl"
+    title: "Implement the requested code change and run tests."
+    type: "implementation"
+    assignee: null
+    deps:
+      - "task_research"
+    acceptance:
+      - "Changed files and tests are reported."
+    status: "pending"
+    requires_tools:
+      - "fs_write"
+      - "execute_bash"
+    files:
+      - "src/**"
+      - "tests/**"
+    risk: "medium"
+    retry_limit: 1
+    human_confirmation: false
+  - id: "task_review"
+    title: "Review implementation output and identify regressions."
+    type: "review"
+    assignee: null
+    deps:
+      - "task_impl"
+    acceptance:
+      - "Findings are structured with risk and artifacts."
+    status: "pending"
+    requires_tools:
+      - "fs_read"
+      - "git_diff"
+    files:
+      - "**/*"
+    risk: "medium"
+    retry_limit: 0
+    human_confirmation: false

package/crates/team-agent/src/model/testdata/team.spec.yaml

@@ -0,0 +1,206 @@
+version: 1
+team:
+  name: teamspec-full-example
+  mode: supervisor_worker
+  objective: Build, research, review, and document a code change with Codex CLI workers.
+  workspace: .
+leader:
+  id: leader
+  role: leader
+  provider: codex
+  model: null
+  tools:
+    - fs_read
+    - fs_list
+    - mcp_team
+    - provider_builtin
+  context_policy:
+    keep_user_thread: true
+    receive_worker_outputs: structured_only
+    max_worker_result_tokens: 4000
+agents:
+  - id: codex_implementer
+    role: implementation_engineer
+    provider: codex
+    model: null
+    working_directory: .
+    system_prompt:
+      inline: |
+        You are the implementation worker. Make focused code changes, run relevant tests,
+        and report a result_envelope_v1 with changed files, tests, risks, artifacts, and next actions.
+      file: null
+    tools:
+      - fs_read
+      - fs_write
+      - fs_list
+      - execute_bash
+      - git_diff
+      - mcp_team
+      - provider_builtin
+    permission_mode: restricted
+    preferred_for:
+      - implementation
+      - bug_fix
+      - test
+    avoid_for:
+      - final_risk_signoff
+    output_contract:
+      format: result_envelope_v1
+      required_fields:
+        - task_id
+        - status
+        - summary
+        - artifacts
+  - id: codex_researcher
+    role: researcher
+    provider: codex
+    model: null
+    working_directory: .
+    system_prompt:
+      inline: |
+        You are the research worker. Prefer read-only analysis and summarize findings
+        as result_envelope_v1. Do not edit files.
+      file: null
+    tools:
+      - fs_read
+      - fs_list
+      - network
+      - mcp_team
+      - provider_builtin
+    permission_mode: restricted
+    preferred_for:
+      - research
+      - architecture
+      - docs
+    avoid_for:
+      - implementation
+    output_contract:
+      format: result_envelope_v1
+      required_fields:
+        - task_id
+        - status
+        - summary
+        - artifacts
+  - id: codex_reviewer
+    role: code_reviewer
+    provider: codex
+    model: null
+    working_directory: .
+    system_prompt:
+      inline: |
+        You are the reviewer. Find correctness, regression, security, and missing-test risks.
+        Stay read-only unless the leader explicitly changes your permissions.
+      file: null
+    tools:
+      - fs_read
+      - fs_list
+      - git_diff
+      - mcp_team
+      - provider_builtin
+    permission_mode: restricted
+    preferred_for:
+      - review
+      - risk_check
+    avoid_for:
+      - implementation
+    output_contract:
+      format: result_envelope_v1
+      required_fields:
+        - task_id
+        - status
+        - summary
+        - artifacts
+routing:
+  default_assignee: leader
+  rules:
+    - id: implementation-to-codex
+      when: task.type in ["implementation", "bug_fix", "test"]
+      assign_to: codex_implementer
+      priority: 100
+    - id: research-to-codex
+      when: task.type in ["research", "architecture", "docs"]
+      assign_to: codex_researcher
+      priority: 90
+    - id: review-to-codex
+      when: task.type in ["review", "risk_check"]
+      assign_to: codex_reviewer
+      priority: 90
+communication:
+  protocol: mcp_inbox
+  topology: leader_centered
+  worker_to_worker: false
+  ack_timeout_sec: 60
+  result_format: result_envelope_v1
+  message_store:
+    sqlite: .team/runtime/team.db
+    mirror_files: .team/messages
+runtime:
+  backend: tmux
+  display_backend: none
+  session_name: teamspec-full-example
+  auto_launch: true
+  require_user_approval_before_launch: true
+  dangerous_auto_approve: false
+  max_active_agents: 3
+  startup_order:
+    - codex_implementer
+    - codex_researcher
+    - codex_reviewer
+context:
+  state_file: team_state.md
+  artifact_dir: .team/artifacts
+  log_dir: .team/logs
+  summarization:
+    worker_full_logs: retain_outside_leader_context
+    state_update: after_each_result
+tasks:
+  - id: task_research
+    title: Read the task context and identify design risks.
+    type: research
+    assignee: null
+    deps: []
+    acceptance:
+      - Result envelope includes summary and risks.
+    status: pending
+    requires_tools:
+      - fs_read
+    files:
+      - "**/*"
+    risk: medium
+    retry_limit: 1
+    human_confirmation: false
+  - id: task_impl
+    title: Implement the requested code change and run tests.
+    type: implementation
+    assignee: null
+    deps:
+      - task_research
+    acceptance:
+      - Changed files and tests are reported.
+    status: pending
+    requires_tools:
+      - fs_write
+      - execute_bash
+    files:
+      - "src/**"
+      - "tests/**"
+    risk: medium
+    retry_limit: 1
+    human_confirmation: false
+  - id: task_review
+    title: Review implementation output and identify regressions.
+    type: review
+    assignee: null
+    deps:
+      - task_impl
+    acceptance:
+      - Findings are structured with risk and artifacts.
+    status: pending
+    requires_tools:
+      - fs_read
+      - git_diff
+    files:
+      - "**/*"
+    risk: medium
+    retry_limit: 0
+    human_confirmation: false

package/crates/team-agent/src/model/yaml/tests.rs

@@ -0,0 +1,288 @@
+#![allow(clippy::unwrap_used)]
+use super::*;
+
+/// 把一段 YAML 文本经 `dumps(loads(t))` 往返,断言**逐字节**等于 Python golden。
+/// 这是 §4.2 双跑对该方言的字节锁:golden 由
+/// `PYTHONPATH=src python3 -c "from team_agent.simple_yaml import loads,dumps; ..."` 取得。
+fn roundtrip_eq(input: &str, golden: &str) {
+    let v = loads(input).unwrap();
+    assert_eq!(dumps(&v), golden, "roundtrip mismatch for input:\n{input}");
+}
+
+// === 真实语料:examples/team.spec.yaml 的 dumps(loads(t)) golden(逐字节) ===
+#[test]
+fn real_spec_yaml_roundtrip_matches_python_golden() {
+    let input = include_str!(concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/src/model/testdata/team.spec.yaml"
+    ));
+    let golden = include_str!(concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/src/model/testdata/team.spec.golden.yaml"
+    ));
+    let v = loads(input).unwrap();
+    assert_eq!(dumps(&v), golden);
+}
+
+// === 富边角 fixture:覆盖 ast.literal_eval 半成功(true/null 非法 → 整列退字符串)、
+//     深层嵌套 map、list-of-map 含内嵌 list/map、block scalar 含空行 等 ===
+#[test]
+fn rich_edge_fixture_roundtrip_matches_python_golden() {
+    let input = include_str!(concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/src/model/testdata/fuzz.yaml"
+    ));
+    let golden = include_str!(concat!(
+        env!("CARGO_MANIFEST_DIR"),
+        "/src/model/testdata/fuzz.golden.yaml"
+    ));
+    let v = loads(input).unwrap();
+    assert_eq!(dumps(&v), golden);
+}
+
+// === SKILL.md front-matter 块(`---` 之间那段)的 loads/dumps ===
+#[test]
+fn skill_front_matter_roundtrip() {
+    // 取自 skills/team-agent/SKILL.md 的 front matter。
+    let input = "name: team-agent\ndescription: Use only when the user explicitly asks to start, operate, inspect, shutdown, or restart a Team Agent team. Treat the team-agent CLI as a sealed appliance.\n";
+    let golden = "name: \"team-agent\"\ndescription: \"Use only when the user explicitly asks to start, operate, inspect, shutdown, or restart a Team Agent team. Treat the team-agent CLI as a sealed appliance.\"\n";
+    roundtrip_eq(input, golden);
+}
+
+// === 方言怪癖逐一对拍(每个 golden 来自 Python 真相源)===
+
+#[test]
+fn block_scalar_basic() {
+    roundtrip_eq("msg: |\n  line one\n  line two\n", "msg: |\n  line one\n  line two\n");
+}
+
+#[test]
+fn block_scalar_blank_line_becomes_padded() {
+    // 空行在 dumps 时产生仅含 pad 的行(尾随空格)—— Python 行为,必须复刻。
+    roundtrip_eq("msg: |\n  a\n\n  b\n", "msg: |\n  a\n  \n  b\n");
+}
+
+#[test]
+fn inline_unquoted_list_stays_string() {
+    // ast.literal_eval('[a, b, c]') 失败 → 原样字符串。
+    roundtrip_eq("tools: [a, b, c]\n", "tools: \"[a, b, c]\"\n");
+}
+
+#[test]
+fn inline_quoted_list_becomes_list() {
+    roundtrip_eq("x: ['a', 'b']\n", "x:\n  - \"a\"\n  - \"b\"\n");
+}
+
+#[test]
+fn ints_canonicalize() {
+    // 007 -> 7,-5 保号。
+    roundtrip_eq("a: 1\nb: -5\nc: 007\n", "a: 1\nb: -5\nc: 7\n");
+}
+
+#[test]
+fn int_plus_and_underscore() {
+    roundtrip_eq("a: +3\nb: 1_000\n", "a: 3\nb: 1000\n");
+}
+
+#[test]
+fn non_ints_stay_strings() {
+    roundtrip_eq("a: 1.5\nb: 0x1f\nc: 1e3\nd: _1\n", "a: \"1.5\"\nb: \"0x1f\"\nc: \"1e3\"\nd: \"_1\"\n");
+}
+
+#[test]
+fn bools_and_nulls() {
+    roundtrip_eq("a: true\nb: False\nc: NULL\nd: ~\n", "a: true\nb: false\nc: null\nd: null\n");
+}
+
+#[test]
+fn empty_collections() {
+    roundtrip_eq("a: {}\nb: []\n", "a: {}\nb: []\n");
+}
+
+#[test]
+fn quoted_strings_unwrap() {
+    roundtrip_eq("a: \"hello\"\nb: 'world'\n", "a: \"hello\"\nb: \"world\"\n");
+}
+
+#[test]
+fn nested_map() {
+    roundtrip_eq("a:\n  b:\n    c: 1\n", "a:\n  b:\n    c: 1\n");
+}
+
+#[test]
+fn list_of_maps() {
+    roundtrip_eq(
+        "items:\n  - id: x\n    val: 1\n  - id: y\n    val: 2\n",
+        "items:\n  - id: \"x\"\n    val: 1\n  - id: \"y\"\n    val: 2\n",
+    );
+}
+
+#[test]
+fn scalar_list_mixed() {
+    roundtrip_eq("a:\n  - 1\n  - two\n  - true\n", "a:\n  - 1\n  - \"two\"\n  - true\n");
+}
+
+#[test]
+fn empty_value_is_null() {
+    roundtrip_eq("a:\nb: 2\n", "a: null\nb: 2\n");
+}
+
+#[test]
+fn inline_comment_not_stripped() {
+    // 整行注释被跳过;行内 `#` 不被当注释,留在值里。
+    roundtrip_eq("# top\na: 1  # inline?\nb: 2\n", "a: \"1  # inline?\"\nb: 2\n");
+}
+
+#[test]
+fn key_no_space_after_colon() {
+    roundtrip_eq("a:1\n", "a: 1\n");
+}
+
+#[test]
+fn special_chars_json_escaped() {
+    roundtrip_eq("a: he said \"hi\"\n", "a: \"he said \\\"hi\\\"\"\n");
+}
+
+#[test]
+fn unicode_preserved() {
+    roundtrip_eq("name: café ☕\n", "name: \"café ☕\"\n");
+}
+
+#[test]
+fn list_item_value_has_colon_and_space() {
+    // foo: bar baz —— item_text 看起来是 key-value(key=foo 合法),value = "bar baz"。
+    roundtrip_eq("a:\n  - foo: bar baz\n", "a:\n  - foo: \"bar baz\"\n");
+}
+
+// === JSON 顶层路径(`{`/`[` 开头 → json.loads,保留键序)===
+
+#[test]
+fn json_object_path_preserves_key_order() {
+    roundtrip_eq("{\"a\": 1, \"b\": [1,2]}", "a: 1\nb:\n  - 1\n  - 2\n");
+}
+
+#[test]
+fn json_array_path() {
+    roundtrip_eq("[1, 2, 3]", "- 1\n- 2\n- 3\n");
+}
+
+#[test]
+fn json_leading_whitespace() {
+    roundtrip_eq("   {\"a\": 1}", "a: 1\n");
+}
+
+#[test]
+fn json_float_becomes_quoted_string() {
+    // json.loads -> float 1.5; dumps -> json.dumps(str(1.5)) = "1.5"。
+    roundtrip_eq("{\"a\": 1.5}", "a: \"1.5\"\n");
+}
+
+#[test]
+fn json_string_with_newline_becomes_block_scalar() {
+    roundtrip_eq("{\"x\": \"with\\nnewline\"}", "x: |\n  with\n  newline\n");
+}
+
+#[test]
+fn json_empty_object_and_array_top_level() {
+    // 顶层 {} / [] → _dump 返回空 → "\n"。
+    roundtrip_eq("{}", "\n");
+    roundtrip_eq("[]", "\n");
+}
+
+// === dumps-only 结构(直接构造 Value,golden 来自 Python dumps)===
+
+#[test]
+fn dump_nested_list_in_list() {
+    let v = Value::Map(vec![(
+        "a".to_string(),
+        Value::List(vec![
+            Value::List(vec![Value::Int(1), Value::Int(2)]),
+            Value::List(vec![Value::Int(3)]),
+        ]),
+    )]);
+    assert_eq!(dumps(&v), "a:\n  -\n    - 1\n    - 2\n  -\n    - 3\n");
+}
+
+#[test]
+fn dump_mixed_list() {
+    let v = Value::List(vec![
+        Value::Int(1),
+        Value::Str("two".to_string()),
+        Value::List(vec![Value::Int(3), Value::Int(4)]),
+        Value::Map(vec![("k".to_string(), Value::Str("v".to_string()))]),
+        Value::Map(Vec::new()),
+        Value::List(Vec::new()),
+    ]);
+    assert_eq!(dumps(&v), "- 1\n- \"two\"\n-\n  - 3\n  - 4\n- k: \"v\"\n- {}\n-\n");
+}
+
+#[test]
+fn dump_multiline_string_in_list_item_is_not_block_scalar() {
+    // 关键不对称:list-of-dict 上下文**不**对多行字符串走 `|`。
+    let v = Value::Map(vec![(
+        "items".to_string(),
+        Value::List(vec![Value::Map(vec![(
+            "prompt".to_string(),
+            Value::Str("line1\nline2".to_string()),
+        )])]),
+    )]);
+    assert_eq!(dumps(&v), "items:\n  - prompt: \"line1\\nline2\"\n");
+}
+
+#[test]
+fn dump_deep_map_in_list_uses_indent_plus_4() {
+    let v = Value::Map(vec![(
+        "items".to_string(),
+        Value::List(vec![Value::Map(vec![
+            (
+                "a".to_string(),
+                Value::Map(vec![("b".to_string(), Value::Int(1))]),
+            ),
+            ("c".to_string(), Value::Int(2)),
+        ])]),
+    )]);
+    assert_eq!(dumps(&v), "items:\n  - a:\n      b: 1\n    c: 2\n");
+}
+
+#[test]
+fn dump_string_that_looks_like_int_or_bool_is_quoted() {
+    let v = Value::Map(vec![
+        ("a".to_string(), Value::Str("5".to_string())),
+        ("b".to_string(), Value::Str("true".to_string())),
+    ]);
+    assert_eq!(dumps(&v), "a: \"5\"\nb: \"true\"\n");
+}
+
+#[test]
+fn dump_string_single_newline_only() {
+    // "\n".rstrip("\n") = "" -> splitlines() = [] -> 只有 "a: |"。
+    let v = Value::Map(vec![("a".to_string(), Value::Str("\n".to_string()))]);
+    assert_eq!(dumps(&v), "a: |\n");
+}
+
+// === loads 错误路径(与 Python raise 对齐)===
+
+#[test]
+fn unexpected_indentation_errors() {
+    assert!(loads("a: 1\n  b: 2\n").is_err());
+}
+
+#[test]
+fn missing_colon_errors() {
+    assert!(loads("just a string\n").is_err());
+}
+
+#[test]
+fn json_escape_matches_python_for_control_and_specials() {
+    // 锁住 serde_json 字符串转义 == Python json.dumps(ensure_ascii=False)。
+    assert_eq!(json_quote("a\"b"), "\"a\\\"b\"");
+    assert_eq!(json_quote("a\\b"), "\"a\\\\b\"");
+    assert_eq!(json_quote("a\nb"), "\"a\\nb\"");
+    assert_eq!(json_quote("tab\there"), "\"tab\\there\"");
+    assert_eq!(json_quote("a\rb"), "\"a\\rb\"");
+    assert_eq!(json_quote("café"), "\"café\"");
+    assert_eq!(json_quote("\u{0001}ctrl"), "\"\\u0001ctrl\"");
+    assert_eq!(json_quote("slash/here"), "\"slash/here\"");
+    assert_eq!(json_quote("back\u{0008}space"), "\"back\\bspace\"");
+    assert_eq!(json_quote("form\u{000c}feed"), "\"form\\ffeed\"");
+}