claude-dev-env 8.111.1 → 8.111.3
This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
- package/hooks/blocking/pr_lifecycle_skill_gate.py +45 -4
- package/hooks/blocking/test_chat_reply_gate.py +22 -18
- package/hooks/blocking/test_pr_lifecycle_skill_gate.py +104 -0
- package/hooks/blocking/test_reply_length_gate.py +45 -35
- package/hooks/features/conduct-gates.md +1 -1
- package/hooks/hooks_constants/pr_lifecycle_skill_gate_constants.py +32 -2
- package/hooks/hooks_constants/reply_length_gate_constants.py +3 -3
- package/hooks/hooks_constants/test_reply_length_gate_constants.py +2 -2
- package/package.json +1 -1
- package/scripts/test_durable_post_lint.py +10 -8
|
@@ -18,6 +18,8 @@ from hooks_constants.pr_lifecycle_skill_gate_constants import (
|
|
|
18
18
|
ALL_BUILD_EVAL_SKILL_NAMES,
|
|
19
19
|
ALL_API_ACTION_NAMES,
|
|
20
20
|
ALL_COMMAND_PREFIX_WORDS,
|
|
21
|
+
ALL_GH_PULL_REQUEST_READ_SUBCOMMANDS,
|
|
22
|
+
ALL_GH_PULL_REQUEST_SUBCOMMANDS,
|
|
21
23
|
ALL_GITHUB_MCP_TOOL_SUFFIXES,
|
|
22
24
|
ALL_PIECE_REPLACEMENT_TEXTS,
|
|
23
25
|
ALL_PREFIX_OPTIONS_WITH_VALUE,
|
|
@@ -28,6 +30,9 @@ from hooks_constants.pr_lifecycle_skill_gate_constants import (
|
|
|
28
30
|
BUILD_EVAL_ARGUMENT_WORD,
|
|
29
31
|
BUILD_EVAL_COMMAND_MARKER,
|
|
30
32
|
COMMAND_PIECE_PATTERN,
|
|
33
|
+
BACKTICK_SUBSTITUTION_GROUP,
|
|
34
|
+
COMMAND_SUBSTITUTION_PATTERN,
|
|
35
|
+
DOLLAR_SUBSTITUTION_GROUP,
|
|
31
36
|
COMMAND_WHITESPACE,
|
|
32
37
|
DENY_DECISION,
|
|
33
38
|
DENY_REASON,
|
|
@@ -48,6 +53,10 @@ from hooks_constants.pr_lifecycle_skill_gate_constants import (
|
|
|
48
53
|
PATH_SEPARATOR_PATTERN,
|
|
49
54
|
PERMISSION_DECISION_REASON_KEY,
|
|
50
55
|
PULL_REQUEST_SCRIPT_NAME,
|
|
56
|
+
ALL_NESTED_PROGRAM_NAMES,
|
|
57
|
+
ALL_NESTED_PROGRAM_WRAPPER_NAMES,
|
|
58
|
+
EVAL_ARGUMENT_SEPARATOR,
|
|
59
|
+
EVAL_PROGRAM_NAME,
|
|
51
60
|
PYTHON_OPTIONS_WITH_VALUE,
|
|
52
61
|
REMOVED_TEXT,
|
|
53
62
|
SEGMENT_BREAK,
|
|
@@ -185,10 +194,20 @@ def _is_git_action(all_words: list[str]) -> bool:
|
|
|
185
194
|
return action in GIT_ACTIONS
|
|
186
195
|
|
|
187
196
|
|
|
197
|
+
def _is_pull_request_action(all_gh_words: list[str]) -> bool:
|
|
198
|
+
all_subcommand_words = all_gh_words[1:]
|
|
199
|
+
if not all_subcommand_words:
|
|
200
|
+
return False
|
|
201
|
+
subcommand, *all_later_words = all_subcommand_words
|
|
202
|
+
if subcommand not in ALL_GH_PULL_REQUEST_READ_SUBCOMMANDS:
|
|
203
|
+
return True
|
|
204
|
+
return any(each_word in ALL_GH_PULL_REQUEST_SUBCOMMANDS for each_word in all_later_words)
|
|
205
|
+
|
|
206
|
+
|
|
188
207
|
def _is_gh_action(all_words: list[str]) -> bool:
|
|
189
208
|
all_gh_words = gh_command_words(all_words[1:])
|
|
190
209
|
if all_gh_words[:1] == [GH_PULL_REQUEST_COMMAND]:
|
|
191
|
-
return
|
|
210
|
+
return _is_pull_request_action(all_gh_words)
|
|
192
211
|
if all_gh_words[:1] != [GH_API_COMMAND]:
|
|
193
212
|
return False
|
|
194
213
|
all_api_arguments = all_words[all_words.index(GH_API_COMMAND, 1) + 1 :]
|
|
@@ -207,7 +226,7 @@ def _words_from_executable(all_words: list[str]) -> list[str]:
|
|
|
207
226
|
if re.match(ENVIRONMENT_ASSIGNMENT_PATTERN, each_word):
|
|
208
227
|
index += 1
|
|
209
228
|
continue
|
|
210
|
-
if each_word not in ALL_COMMAND_PREFIX_WORDS:
|
|
229
|
+
if file_name(each_word) not in ALL_COMMAND_PREFIX_WORDS:
|
|
211
230
|
return all_words[index:]
|
|
212
231
|
index += 1
|
|
213
232
|
while index < len(all_words) and all_words[index].startswith("-"):
|
|
@@ -247,12 +266,34 @@ def executable_word_lists(command: str) -> list[list[str]]:
|
|
|
247
266
|
all_command_words = _words_from_executable(each_segment)
|
|
248
267
|
if not all_command_words:
|
|
249
268
|
continue
|
|
250
|
-
all_word_lists.
|
|
251
|
-
|
|
269
|
+
all_word_lists.extend(_command_and_nested_word_lists(all_command_words))
|
|
270
|
+
for each_match in COMMAND_SUBSTITUTION_PATTERN.finditer(command):
|
|
271
|
+
inner_command = (
|
|
272
|
+
each_match.group(DOLLAR_SUBSTITUTION_GROUP)
|
|
273
|
+
if each_match.group(DOLLAR_SUBSTITUTION_GROUP) is not None
|
|
274
|
+
else each_match.group(BACKTICK_SUBSTITUTION_GROUP)
|
|
275
|
+
)
|
|
276
|
+
all_word_lists.extend(executable_word_lists(inner_command))
|
|
277
|
+
return all_word_lists
|
|
278
|
+
|
|
279
|
+
|
|
280
|
+
def _command_and_nested_word_lists(all_command_words: list[str]) -> list[list[str]]:
|
|
281
|
+
return [all_command_words, *_wrapped_word_lists(all_command_words), *_nested_program_word_lists(all_command_words)]
|
|
282
|
+
|
|
283
|
+
|
|
284
|
+
def _nested_program_word_lists(all_command_words: list[str]) -> list[list[str]]:
|
|
285
|
+
if file_name(all_command_words[0]) not in ALL_NESTED_PROGRAM_WRAPPER_NAMES:
|
|
286
|
+
return []
|
|
287
|
+
all_word_lists: list[list[str]] = []
|
|
288
|
+
for each_index in range(1, len(all_command_words)):
|
|
289
|
+
if file_name(all_command_words[each_index]) in ALL_NESTED_PROGRAM_NAMES:
|
|
290
|
+
all_word_lists.extend(_command_and_nested_word_lists(all_command_words[each_index:]))
|
|
252
291
|
return all_word_lists
|
|
253
292
|
|
|
254
293
|
|
|
255
294
|
def _wrapped_word_lists(all_command_words: list[str]) -> list[list[str]]:
|
|
295
|
+
if file_name(all_command_words[0]) == EVAL_PROGRAM_NAME:
|
|
296
|
+
return executable_word_lists(EVAL_ARGUMENT_SEPARATOR.join(all_command_words[1:]))
|
|
256
297
|
if file_name(all_command_words[0]) not in ALL_SHELL_WRAPPER_EXECUTABLES:
|
|
257
298
|
return []
|
|
258
299
|
return [
|
|
@@ -5,8 +5,14 @@ from pathlib import Path
|
|
|
5
5
|
import pytest
|
|
6
6
|
|
|
7
7
|
import chat_reply_gate
|
|
8
|
-
|
|
9
|
-
|
|
8
|
+
from hooks_constants.reply_length_gate_constants import MAXIMUM_SENTENCE_COUNT
|
|
9
|
+
|
|
10
|
+
OVER_LIMIT_SENTENCES = [
|
|
11
|
+
f"Status line {each_number} is ready." for each_number in range(MAXIMUM_SENTENCE_COUNT + 1)
|
|
12
|
+
]
|
|
13
|
+
OVER_LIMIT_REPLY = " ".join(OVER_LIMIT_SENTENCES)
|
|
14
|
+
FIRST_HELD_BATCH = " ".join(OVER_LIMIT_SENTENCES[: len(OVER_LIMIT_SENTENCES) // 2])
|
|
15
|
+
LAST_HELD_BATCH = " ".join(OVER_LIMIT_SENTENCES[len(OVER_LIMIT_SENTENCES) // 2 :])
|
|
10
16
|
THREE_SENTENCE_REPLY = "PR one merged. PR two is queued. Nothing else is open."
|
|
11
17
|
PASSING_DRAFT = "PR one merged. Two others wait."
|
|
12
18
|
|
|
@@ -46,7 +52,7 @@ def test_should_show_a_passing_draft_in_place_of_a_long_reply(
|
|
|
46
52
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
47
53
|
) -> None:
|
|
48
54
|
monkeypatch.setattr(chat_reply_gate, "ask_drafter", lambda *_args: PASSING_DRAFT)
|
|
49
|
-
exit_code, hook_output = run_hook(monkeypatch, capsys, display_input(
|
|
55
|
+
exit_code, hook_output = run_hook(monkeypatch, capsys, display_input(OVER_LIMIT_REPLY))
|
|
50
56
|
assert exit_code == 0
|
|
51
57
|
assert hook_output == {"hookEventName": "MessageDisplay", "displayContent": PASSING_DRAFT}
|
|
52
58
|
|
|
@@ -70,16 +76,14 @@ def test_should_hold_batches_until_the_last_one_arrives(
|
|
|
70
76
|
lambda reply_text, _violation: all_requests.append(reply_text) or PASSING_DRAFT,
|
|
71
77
|
)
|
|
72
78
|
_, first_output = run_hook(
|
|
73
|
-
monkeypatch, capsys, display_input(
|
|
79
|
+
monkeypatch, capsys, display_input(FIRST_HELD_BATCH + "\n", False)
|
|
74
80
|
)
|
|
75
81
|
_, last_output = run_hook(
|
|
76
|
-
monkeypatch, capsys, display_input(
|
|
82
|
+
monkeypatch, capsys, display_input(LAST_HELD_BATCH)
|
|
77
83
|
)
|
|
78
84
|
assert first_output == {"hookEventName": "MessageDisplay", "displayContent": ""}
|
|
79
85
|
assert last_output == {"hookEventName": "MessageDisplay", "displayContent": PASSING_DRAFT}
|
|
80
|
-
assert all_requests == [
|
|
81
|
-
"PR one merged. PR two is queued.\nPR three needs a fix. Nothing else is open."
|
|
82
|
-
]
|
|
86
|
+
assert all_requests == [FIRST_HELD_BATCH + "\n" + LAST_HELD_BATCH]
|
|
83
87
|
|
|
84
88
|
|
|
85
89
|
def test_should_show_the_whole_held_message_when_it_passes(
|
|
@@ -100,9 +104,9 @@ def test_should_show_the_original_when_no_draft_passes(
|
|
|
100
104
|
monkeypatch.setattr(
|
|
101
105
|
chat_reply_gate,
|
|
102
106
|
"ask_drafter",
|
|
103
|
-
lambda reply_text, _violation: all_calls.append(reply_text) or
|
|
107
|
+
lambda reply_text, _violation: all_calls.append(reply_text) or OVER_LIMIT_REPLY,
|
|
104
108
|
)
|
|
105
|
-
_, hook_output = run_hook(monkeypatch, capsys, display_input(
|
|
109
|
+
_, hook_output = run_hook(monkeypatch, capsys, display_input(OVER_LIMIT_REPLY))
|
|
106
110
|
assert hook_output is None
|
|
107
111
|
assert len(all_calls) == 2
|
|
108
112
|
|
|
@@ -111,7 +115,7 @@ def test_should_tell_the_agent_which_draft_the_user_saw(
|
|
|
111
115
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
112
116
|
) -> None:
|
|
113
117
|
monkeypatch.setattr(chat_reply_gate, "ask_drafter", lambda *_args: PASSING_DRAFT)
|
|
114
|
-
run_hook(monkeypatch, capsys, display_input(
|
|
118
|
+
run_hook(monkeypatch, capsys, display_input(OVER_LIMIT_REPLY))
|
|
115
119
|
prompt_input = {"hook_event_name": "UserPromptSubmit", "session_id": "abc"}
|
|
116
120
|
_, first_output = run_hook(monkeypatch, capsys, prompt_input)
|
|
117
121
|
_, second_output = run_hook(monkeypatch, capsys, prompt_input)
|
|
@@ -124,7 +128,7 @@ def test_should_delete_the_shown_draft_at_session_end(
|
|
|
124
128
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
125
129
|
) -> None:
|
|
126
130
|
monkeypatch.setattr(chat_reply_gate, "ask_drafter", lambda *_args: PASSING_DRAFT)
|
|
127
|
-
run_hook(monkeypatch, capsys, display_input(
|
|
131
|
+
run_hook(monkeypatch, capsys, display_input(OVER_LIMIT_REPLY))
|
|
128
132
|
shown_draft_path = chat_reply_gate.state_file_path("claude-chat-reply-shown-", "abc")
|
|
129
133
|
assert shown_draft_path.exists()
|
|
130
134
|
exit_code, hook_output = run_hook(
|
|
@@ -144,14 +148,14 @@ def test_should_add_the_reply_limit_on_each_user_prompt(
|
|
|
144
148
|
assert exit_code == 0
|
|
145
149
|
assert hook_output is not None
|
|
146
150
|
assert hook_output["hookEventName"] == "UserPromptSubmit"
|
|
147
|
-
assert "
|
|
151
|
+
assert f"{MAXIMUM_SENTENCE_COUNT} sentences or fewer" in str(hook_output["additionalContext"])
|
|
148
152
|
|
|
149
153
|
|
|
150
154
|
def test_should_do_nothing_inside_the_drafter_process(
|
|
151
155
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
152
156
|
) -> None:
|
|
153
157
|
monkeypatch.setenv("CHAT_REPLY_DRAFTER_ACTIVE", "1")
|
|
154
|
-
exit_code, hook_output = run_hook(monkeypatch, capsys, display_input(
|
|
158
|
+
exit_code, hook_output = run_hook(monkeypatch, capsys, display_input(OVER_LIMIT_REPLY))
|
|
155
159
|
assert exit_code == 0
|
|
156
160
|
assert hook_output is None
|
|
157
161
|
|
|
@@ -162,7 +166,7 @@ def test_should_ignore_the_stop_event(
|
|
|
162
166
|
exit_code, hook_output = run_hook(
|
|
163
167
|
monkeypatch,
|
|
164
168
|
capsys,
|
|
165
|
-
{"hook_event_name": "Stop", "last_assistant_message":
|
|
169
|
+
{"hook_event_name": "Stop", "last_assistant_message": OVER_LIMIT_REPLY},
|
|
166
170
|
)
|
|
167
171
|
assert exit_code == 0
|
|
168
172
|
assert hook_output is None
|
|
@@ -179,7 +183,7 @@ def test_ask_drafter_should_read_the_result_field(monkeypatch: pytest.MonkeyPatc
|
|
|
179
183
|
"run",
|
|
180
184
|
lambda command, **_kwargs: all_commands.append(command) or CompletedDrafter(),
|
|
181
185
|
)
|
|
182
|
-
assert chat_reply_gate.ask_drafter(
|
|
186
|
+
assert chat_reply_gate.ask_drafter(OVER_LIMIT_REPLY, "too long") == PASSING_DRAFT
|
|
183
187
|
assert all_commands[0][:4] == ["C:/npm/claude.cmd", "-p", "--model", "haiku"]
|
|
184
188
|
|
|
185
189
|
|
|
@@ -188,14 +192,14 @@ def test_ask_drafter_should_return_none_for_an_error_result(monkeypatch: pytest.
|
|
|
188
192
|
stdout = json.dumps({"is_error": True, "result": "Credit balance is too low"})
|
|
189
193
|
|
|
190
194
|
monkeypatch.setattr(chat_reply_gate.subprocess, "run", lambda *_args, **_kwargs: FailedDrafter())
|
|
191
|
-
assert chat_reply_gate.ask_drafter(
|
|
195
|
+
assert chat_reply_gate.ask_drafter(OVER_LIMIT_REPLY, "too long") is None
|
|
192
196
|
|
|
193
197
|
|
|
194
198
|
def test_should_forget_the_shown_draft_when_a_later_reply_passes(
|
|
195
199
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
196
200
|
) -> None:
|
|
197
201
|
monkeypatch.setattr(chat_reply_gate, "ask_drafter", lambda *_args: PASSING_DRAFT)
|
|
198
|
-
run_hook(monkeypatch, capsys, display_input(
|
|
202
|
+
run_hook(monkeypatch, capsys, display_input(OVER_LIMIT_REPLY))
|
|
199
203
|
run_hook(monkeypatch, capsys, {**display_input(THREE_SENTENCE_REPLY), "message_id": "m2"})
|
|
200
204
|
_, hook_output = run_hook(monkeypatch, capsys, {"hook_event_name": "UserPromptSubmit", "session_id": "abc"})
|
|
201
205
|
assert hook_output is not None
|
|
@@ -510,3 +510,107 @@ def test_deny_decision_should_carry_the_reason_as_a_deny() -> None:
|
|
|
510
510
|
output = gate.deny_decision("Load the skill first.")["hookSpecificOutput"]
|
|
511
511
|
assert output["permissionDecision"] == "deny"
|
|
512
512
|
assert output["permissionDecisionReason"] == "Load the skill first."
|
|
513
|
+
|
|
514
|
+
|
|
515
|
+
READ_ONLY_PULL_REQUEST_COMMANDS = (
|
|
516
|
+
"gh pr list",
|
|
517
|
+
"gh pr list --state all",
|
|
518
|
+
"gh pr view 12",
|
|
519
|
+
"gh pr view 12 --web",
|
|
520
|
+
"gh pr checks 12",
|
|
521
|
+
"gh pr diff 12",
|
|
522
|
+
"gh pr status",
|
|
523
|
+
"gh -R jl-cmd-projects/claude-dev-env pr view 12",
|
|
524
|
+
)
|
|
525
|
+
WRITE_PULL_REQUEST_COMMANDS = (
|
|
526
|
+
"gh pr create --title t --body b",
|
|
527
|
+
"gh pr merge 12 --squash",
|
|
528
|
+
"gh pr edit 12 --body b",
|
|
529
|
+
"gh pr comment 12 --body b",
|
|
530
|
+
"gh pr review 12 --approve",
|
|
531
|
+
"gh pr close 12",
|
|
532
|
+
"gh pr ready 12",
|
|
533
|
+
"gh pr frobnicate 12",
|
|
534
|
+
)
|
|
535
|
+
|
|
536
|
+
|
|
537
|
+
def test_read_only_pull_request_commands_should_pass_without_the_skill(tmp_path: Path) -> None:
|
|
538
|
+
path = _transcript(tmp_path)
|
|
539
|
+
for each_command in READ_ONLY_PULL_REQUEST_COMMANDS:
|
|
540
|
+
assert gate.decision_for(_payload(each_command, path)) is None, each_command
|
|
541
|
+
|
|
542
|
+
|
|
543
|
+
def test_read_only_pull_request_command_should_run_silently_without_the_skill(tmp_path: Path) -> None:
|
|
544
|
+
assert _run_main(_payload("gh pr list", _transcript(tmp_path))) == (0, "")
|
|
545
|
+
|
|
546
|
+
|
|
547
|
+
def test_write_pull_request_commands_should_be_denied_without_the_skill(tmp_path: Path) -> None:
|
|
548
|
+
path = _transcript(tmp_path)
|
|
549
|
+
for each_command in WRITE_PULL_REQUEST_COMMANDS:
|
|
550
|
+
_assert_denied(_payload(each_command, path))
|
|
551
|
+
|
|
552
|
+
|
|
553
|
+
OPTION_SHIFTED_WRITE_COMMANDS = (
|
|
554
|
+
"gh pr --subject view merge 12 --squash",
|
|
555
|
+
"gh pr --subject view 12 merge",
|
|
556
|
+
)
|
|
557
|
+
|
|
558
|
+
|
|
559
|
+
def test_write_behind_an_unlisted_option_should_be_denied_without_the_skill(tmp_path: Path) -> None:
|
|
560
|
+
path = _transcript(tmp_path)
|
|
561
|
+
for each_command in OPTION_SHIFTED_WRITE_COMMANDS:
|
|
562
|
+
_assert_denied(_payload(each_command, path))
|
|
563
|
+
|
|
564
|
+
|
|
565
|
+
XARGS_WRITE_COMMANDS = (
|
|
566
|
+
"gh pr view 12 | xargs gh pr merge",
|
|
567
|
+
"echo 12 | xargs -I {} gh pr merge {}",
|
|
568
|
+
"echo 12 | xargs -n 1 -P 4 gh pr merge",
|
|
569
|
+
"echo 12 | xargs -- gh pr merge",
|
|
570
|
+
"echo https://example.test | xargs git push origin HEAD",
|
|
571
|
+
)
|
|
572
|
+
XARGS_READ_COMMANDS = (
|
|
573
|
+
"echo 12 | xargs gh pr view",
|
|
574
|
+
"echo 12 | xargs -I {} gh pr checks {}",
|
|
575
|
+
)
|
|
576
|
+
|
|
577
|
+
|
|
578
|
+
def test_write_run_through_xargs_should_be_denied_without_the_skill(tmp_path: Path) -> None:
|
|
579
|
+
path = _transcript(tmp_path)
|
|
580
|
+
for each_command in XARGS_WRITE_COMMANDS:
|
|
581
|
+
_assert_denied(_payload(each_command, path))
|
|
582
|
+
|
|
583
|
+
|
|
584
|
+
def test_read_only_run_through_xargs_should_pass_without_the_skill(tmp_path: Path) -> None:
|
|
585
|
+
path = _transcript(tmp_path)
|
|
586
|
+
for each_command in XARGS_READ_COMMANDS:
|
|
587
|
+
assert gate.decision_for(_payload(each_command, path)) is None, each_command
|
|
588
|
+
|
|
589
|
+
|
|
590
|
+
WRAPPED_WRITE_COMMANDS = (
|
|
591
|
+
"timeout 10 gh pr merge 12",
|
|
592
|
+
"nice -n 5 gh pr merge 12",
|
|
593
|
+
"find . -maxdepth 1 -exec gh pr merge 12 {} ;",
|
|
594
|
+
"parallel gh pr merge 12 ::: 12",
|
|
595
|
+
'eval "gh pr merge 12"',
|
|
596
|
+
"eval gh pr merge 12",
|
|
597
|
+
'eval "gh pr" "merge 12"',
|
|
598
|
+
)
|
|
599
|
+
SUBSTITUTED_AND_GRAPHQL_WRITE_COMMANDS = (
|
|
600
|
+
'echo "$(gh pr merge 12 --squash)"',
|
|
601
|
+
"echo `gh pr merge 12`",
|
|
602
|
+
"/usr/bin/env gh pr merge 12",
|
|
603
|
+
"gh api graphql -f query='mutation { mergePullRequest(input: {pullRequestId: \"PR_x\"}) { clientMutationId } }'",
|
|
604
|
+
)
|
|
605
|
+
|
|
606
|
+
|
|
607
|
+
def test_substituted_or_graphql_pull_request_write_should_be_denied_without_the_skill(tmp_path: Path) -> None:
|
|
608
|
+
path = _transcript(tmp_path)
|
|
609
|
+
for each_command in SUBSTITUTED_AND_GRAPHQL_WRITE_COMMANDS:
|
|
610
|
+
_assert_denied(_payload(each_command, path))
|
|
611
|
+
|
|
612
|
+
|
|
613
|
+
def test_write_run_through_a_wrapper_should_be_denied_without_the_skill(tmp_path: Path) -> None:
|
|
614
|
+
path = _transcript(tmp_path)
|
|
615
|
+
for each_command in WRAPPED_WRITE_COMMANDS:
|
|
616
|
+
_assert_denied(_payload(each_command, path))
|
|
@@ -5,15 +5,25 @@ from pathlib import Path
|
|
|
5
5
|
import pytest
|
|
6
6
|
|
|
7
7
|
import reply_length_gate
|
|
8
|
+
from hooks_constants.reply_length_gate_constants import (
|
|
9
|
+
MAXIMUM_SENTENCE_COUNT,
|
|
10
|
+
MAXIMUM_WORDS_PER_SENTENCE,
|
|
11
|
+
)
|
|
8
12
|
|
|
9
13
|
REPLY_TOOL_NAME = "mcp__hearthbot__reply"
|
|
10
14
|
POST_TOOL_NAME = "mcp__hearthbot__post_message"
|
|
11
|
-
|
|
12
|
-
"
|
|
13
|
-
|
|
15
|
+
OVER_SENTENCE_LIMIT_TEXT = " ".join(
|
|
16
|
+
f"Status line {each_number} is ready." for each_number in range(MAXIMUM_SENTENCE_COUNT + 1)
|
|
17
|
+
)
|
|
18
|
+
OVER_WORD_LIMIT_SENTENCE = (
|
|
19
|
+
"One two three "
|
|
20
|
+
+ " ".join(f"word{each_number}" for each_number in range(3, MAXIMUM_WORDS_PER_SENTENCE + 1))
|
|
21
|
+
+ "."
|
|
14
22
|
)
|
|
15
|
-
|
|
16
|
-
"One two three
|
|
23
|
+
AT_WORD_LIMIT_SENTENCE = (
|
|
24
|
+
"One two three "
|
|
25
|
+
+ " ".join(f"word{each_number}" for each_number in range(3, MAXIMUM_WORDS_PER_SENTENCE))
|
|
26
|
+
+ "."
|
|
17
27
|
)
|
|
18
28
|
|
|
19
29
|
|
|
@@ -59,43 +69,43 @@ def test_should_allow_three_short_sentences(
|
|
|
59
69
|
assert (exit_code, stderr_text) == (0, "")
|
|
60
70
|
|
|
61
71
|
|
|
62
|
-
def
|
|
72
|
+
def test_should_allow_a_sentence_at_the_word_limit(
|
|
63
73
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
64
74
|
) -> None:
|
|
65
|
-
exit_code, _ = run_gate(monkeypatch, capsys, REPLY_TOOL_NAME, {"text":
|
|
75
|
+
exit_code, _ = run_gate(monkeypatch, capsys, REPLY_TOOL_NAME, {"text": AT_WORD_LIMIT_SENTENCE})
|
|
66
76
|
assert exit_code == 0
|
|
67
77
|
|
|
68
78
|
|
|
69
|
-
def
|
|
79
|
+
def test_should_deny_too_many_sentences_and_name_the_count(
|
|
70
80
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
71
81
|
) -> None:
|
|
72
82
|
exit_code, stderr_text = run_gate(
|
|
73
83
|
monkeypatch,
|
|
74
84
|
capsys,
|
|
75
85
|
REPLY_TOOL_NAME,
|
|
76
|
-
{"text":
|
|
86
|
+
{"text": OVER_SENTENCE_LIMIT_TEXT},
|
|
77
87
|
)
|
|
78
88
|
assert exit_code == 2
|
|
79
|
-
assert "
|
|
80
|
-
assert "
|
|
89
|
+
assert f"{MAXIMUM_SENTENCE_COUNT + 1} sentences" in stderr_text
|
|
90
|
+
assert f"limit {MAXIMUM_SENTENCE_COUNT}" in stderr_text
|
|
81
91
|
|
|
82
92
|
|
|
83
|
-
def
|
|
93
|
+
def test_should_deny_a_sentence_over_the_word_limit_and_name_it(
|
|
84
94
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
85
95
|
) -> None:
|
|
86
96
|
exit_code, stderr_text = run_gate(
|
|
87
|
-
monkeypatch, capsys, REPLY_TOOL_NAME, {"text":
|
|
97
|
+
monkeypatch, capsys, REPLY_TOOL_NAME, {"text": OVER_WORD_LIMIT_SENTENCE}
|
|
88
98
|
)
|
|
89
99
|
assert exit_code == 2
|
|
90
|
-
assert "
|
|
91
|
-
assert "
|
|
100
|
+
assert f"{MAXIMUM_WORDS_PER_SENTENCE + 1} words" in stderr_text
|
|
101
|
+
assert f"limit {MAXIMUM_WORDS_PER_SENTENCE}" in stderr_text
|
|
92
102
|
assert "One two three" in stderr_text
|
|
93
103
|
|
|
94
104
|
|
|
95
105
|
def test_should_check_the_project_chat_post_tool(
|
|
96
106
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
97
107
|
) -> None:
|
|
98
|
-
exit_code, _ = run_gate(monkeypatch, capsys, POST_TOOL_NAME, {"text":
|
|
108
|
+
exit_code, _ = run_gate(monkeypatch, capsys, POST_TOOL_NAME, {"text": OVER_SENTENCE_LIMIT_TEXT})
|
|
99
109
|
assert exit_code == 2
|
|
100
110
|
|
|
101
111
|
|
|
@@ -106,10 +116,10 @@ def test_should_count_each_line_of_a_list_as_a_sentence(
|
|
|
106
116
|
monkeypatch,
|
|
107
117
|
capsys,
|
|
108
118
|
REPLY_TOOL_NAME,
|
|
109
|
-
{"text": "
|
|
119
|
+
{"text": "\n".join(f"- item {each_number}" for each_number in range(MAXIMUM_SENTENCE_COUNT + 1))},
|
|
110
120
|
)
|
|
111
121
|
assert exit_code == 2
|
|
112
|
-
assert "
|
|
122
|
+
assert f"{MAXIMUM_SENTENCE_COUNT + 1} sentences" in stderr_text
|
|
113
123
|
|
|
114
124
|
|
|
115
125
|
def test_should_not_count_urls_as_words(
|
|
@@ -156,7 +166,7 @@ def test_should_not_count_backticked_spans_as_words(
|
|
|
156
166
|
def test_should_not_count_fenced_blocks(
|
|
157
167
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
158
168
|
) -> None:
|
|
159
|
-
fenced_block = "```\nOne.\nTwo.\nThree.\nFour.\n" +
|
|
169
|
+
fenced_block = "```\nOne.\nTwo.\nThree.\nFour.\n" + OVER_WORD_LIMIT_SENTENCE + "\n```"
|
|
160
170
|
exit_code, _ = run_gate(
|
|
161
171
|
monkeypatch,
|
|
162
172
|
capsys,
|
|
@@ -369,14 +379,14 @@ def short_card(**all_overrides: object) -> dict[str, object]:
|
|
|
369
379
|
return card
|
|
370
380
|
|
|
371
381
|
|
|
372
|
-
def
|
|
382
|
+
def test_should_allow_a_card_context_of_two_sentences_at_the_word_limit(
|
|
373
383
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
374
384
|
) -> None:
|
|
375
385
|
exit_code, stderr_text = run_gate(
|
|
376
386
|
monkeypatch,
|
|
377
387
|
capsys,
|
|
378
388
|
DECISION_TOOL_NAME,
|
|
379
|
-
short_card(context=" ".join([
|
|
389
|
+
short_card(context=" ".join([AT_WORD_LIMIT_SENTENCE] * 2)),
|
|
380
390
|
)
|
|
381
391
|
assert (exit_code, stderr_text) == (0, "")
|
|
382
392
|
|
|
@@ -439,13 +449,13 @@ def test_should_deny_a_long_card_consequence(
|
|
|
439
449
|
DECISION_TOOL_NAME,
|
|
440
450
|
short_card(
|
|
441
451
|
options=[
|
|
442
|
-
{"label": "Exclude it", "consequence":
|
|
452
|
+
{"label": "Exclude it", "consequence": OVER_WORD_LIMIT_SENTENCE},
|
|
443
453
|
{"label": "Leave it", "consequence": "Seven themes keep a warning."},
|
|
444
454
|
]
|
|
445
455
|
),
|
|
446
456
|
)
|
|
447
457
|
assert exit_code == 2
|
|
448
|
-
assert "Sentence too long:
|
|
458
|
+
assert f"Sentence too long: {MAXIMUM_WORDS_PER_SENTENCE + 1} words, limit {MAXIMUM_WORDS_PER_SENTENCE}" in stderr_text
|
|
449
459
|
|
|
450
460
|
|
|
451
461
|
@pytest.mark.parametrize("card_field_name", ["question", "reason"])
|
|
@@ -456,10 +466,10 @@ def test_should_deny_a_long_sentence_in_the_card_question_and_reason(
|
|
|
456
466
|
monkeypatch,
|
|
457
467
|
capsys,
|
|
458
468
|
DECISION_TOOL_NAME,
|
|
459
|
-
short_card(**{card_field_name:
|
|
469
|
+
short_card(**{card_field_name: OVER_WORD_LIMIT_SENTENCE}),
|
|
460
470
|
)
|
|
461
471
|
assert exit_code == 2
|
|
462
|
-
assert "Sentence too long:
|
|
472
|
+
assert f"Sentence too long: {MAXIMUM_WORDS_PER_SENTENCE + 1} words" in stderr_text
|
|
463
473
|
|
|
464
474
|
|
|
465
475
|
@pytest.mark.parametrize(
|
|
@@ -760,7 +770,7 @@ def test_should_keep_the_length_check_when_the_mode_is_off(
|
|
|
760
770
|
monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str], tmp_path: Path
|
|
761
771
|
) -> None:
|
|
762
772
|
turn_mode_off(tmp_path)
|
|
763
|
-
exit_code, _ = run_gate(monkeypatch, capsys, REPLY_TOOL_NAME, {"text":
|
|
773
|
+
exit_code, _ = run_gate(monkeypatch, capsys, REPLY_TOOL_NAME, {"text": OVER_SENTENCE_LIMIT_TEXT})
|
|
764
774
|
assert exit_code == 2
|
|
765
775
|
|
|
766
776
|
|
|
@@ -1025,7 +1035,7 @@ def test_should_split_prose_sentences_outside_fenced_blocks() -> None:
|
|
|
1025
1035
|
|
|
1026
1036
|
|
|
1027
1037
|
LONG_GITHUB_COMMENT = " ".join(
|
|
1028
|
-
"Sentence number one of a long comment that keeps going." for _ in range(
|
|
1038
|
+
"Sentence number one of a long comment that keeps going." for _ in range(MAXIMUM_SENTENCE_COUNT + 1)
|
|
1029
1039
|
)
|
|
1030
1040
|
SHORT_GITHUB_COMMENT = "Fixed the cap check. The test now covers it. Checks pass."
|
|
1031
1041
|
|
|
@@ -1037,7 +1047,7 @@ def test_should_deny_long_github_comment(
|
|
|
1037
1047
|
monkeypatch, capsys, "mcp__github__add_issue_comment", {"body": LONG_GITHUB_COMMENT}
|
|
1038
1048
|
)
|
|
1039
1049
|
assert exit_code == 2
|
|
1040
|
-
assert "
|
|
1050
|
+
assert f"{MAXIMUM_SENTENCE_COUNT + 1} sentences, limit {MAXIMUM_SENTENCE_COUNT}" in error_text
|
|
1041
1051
|
|
|
1042
1052
|
|
|
1043
1053
|
def test_should_allow_short_github_comment_with_code_and_details(
|
|
@@ -1124,13 +1134,13 @@ def test_should_name_every_broken_rule_in_one_refusal(
|
|
|
1124
1134
|
monkeypatch,
|
|
1125
1135
|
capsys,
|
|
1126
1136
|
POST_TOOL_NAME,
|
|
1127
|
-
{"text": f"See PR 7452 and PR 7460.\n{
|
|
1137
|
+
{"text": f"See PR 7452 and PR 7460.\n{OVER_WORD_LIMIT_SENTENCE}\nIt failed — again."},
|
|
1128
1138
|
)
|
|
1129
1139
|
|
|
1130
1140
|
assert exit_code == reply_length_gate.BLOCK_EXIT_CODE
|
|
1131
1141
|
assert "PR 7452" in error_text
|
|
1132
1142
|
assert "PR 7460" in error_text
|
|
1133
|
-
assert "
|
|
1143
|
+
assert f"{MAXIMUM_WORDS_PER_SENTENCE + 1} words" in error_text
|
|
1134
1144
|
assert reply_length_gate.EM_DASH_MESSAGE in error_text
|
|
1135
1145
|
assert error_text.endswith(reply_length_gate.FIX_EVERY_ITEM_INSTRUCTION)
|
|
1136
1146
|
|
|
@@ -1142,16 +1152,16 @@ def test_should_number_nested_violations_once_in_one_refusal(
|
|
|
1142
1152
|
monkeypatch,
|
|
1143
1153
|
capsys,
|
|
1144
1154
|
POST_TOOL_NAME,
|
|
1145
|
-
{"text": f"
|
|
1155
|
+
{"text": f"{OVER_SENTENCE_LIMIT_TEXT}\n{OVER_WORD_LIMIT_SENTENCE}\n{OVER_WORD_LIMIT_SENTENCE}\nIt failed — again."},
|
|
1146
1156
|
)
|
|
1147
1157
|
|
|
1148
1158
|
assert exit_code == reply_length_gate.BLOCK_EXIT_CODE
|
|
1149
1159
|
all_numbered_lines = [each_line for each_line in error_text.splitlines() if each_line[:1].isdigit()]
|
|
1150
|
-
assert [each_line.split(". ", 1)[0] for each_line in all_numbered_lines] == ["1", "2", "3", "4"]
|
|
1160
|
+
assert [each_line.split(". ", 1)[0] for each_line in all_numbered_lines] == ["1", "2", "3", "4", "5"]
|
|
1151
1161
|
assert not any(each_line.split(". ", 1)[1][:1].isdigit() for each_line in all_numbered_lines)
|
|
1152
|
-
assert "Reply too long:
|
|
1153
|
-
assert all_numbered_lines[1].count("
|
|
1154
|
-
assert all_numbered_lines[2].count("
|
|
1162
|
+
assert f"Reply too long: {MAXIMUM_SENTENCE_COUNT + 4} sentences" in all_numbered_lines[0]
|
|
1163
|
+
assert all_numbered_lines[1].count(f"{MAXIMUM_WORDS_PER_SENTENCE + 1} words") == 1
|
|
1164
|
+
assert all_numbered_lines[2].count(f"{MAXIMUM_WORDS_PER_SENTENCE + 1} words") == 1
|
|
1155
1165
|
assert reply_length_gate.EM_DASH_MESSAGE in all_numbered_lines[3]
|
|
1156
1166
|
|
|
1157
1167
|
|
|
@@ -17,7 +17,7 @@ This family stops a chat reply, edited message, or tool call when its payload br
|
|
|
17
17
|
- `blocking/step_note_gate.py` requires a status line before a tool call while its opt-in flag is on.
|
|
18
18
|
- `blocking/wakeup_cap_gate.py` keeps a self-wakeup inside the 1-hour prompt cache. A wake over the step-down (57 minutes for `send_later` and Routines, 3480 seconds for `ScheduleWakeup`) and at most 60 minutes away runs with its delay rewritten to the step-down. A wake over 60 minutes away, or a cron with more than 57 minutes between wakes, is denied with a reason that says to stack wakes under the cap. A Routine that starts a fresh session on each firing passes.
|
|
19
19
|
- `blocking/verify_before_acting.py` blocks a completed mutating call when the agent's reasoning contains an unchecked claim.
|
|
20
|
-
- `blocking/pr_lifecycle_skill_gate.py` denies a commit, push, pull request action, or merge until the `pr-lifecycle` skill appears in the transcript since the last compaction.
|
|
20
|
+
- `blocking/pr_lifecycle_skill_gate.py` denies a commit, push, pull request action, or merge until the `pr-lifecycle` skill appears in the transcript since the last compaction. Read-only `gh pr` subcommands (`view`, `list`, `checks`, `diff`, `status`) pass without it, unless a later argument names another `gh pr` subcommand.
|
|
21
21
|
- `blocking/resolve_thread_reply_gate.py` denies a review thread resolve until the agent's own transcript shows a reply on that same thread that posted after the last user prompt and that no successful resolve of the thread has spent. A resolve is a `resolve_review_thread` call, which names the thread by `threadID`; a `pull_request_review_write` call with method `resolve_thread`, which names it by `threadId`; a shell command with a `gh api graphql` `resolveReviewThread` mutation, which names it by a literal `PRRT_` id; or a shell call to `pulls/<n>/ccr/comments/<id>/resolve`, which names it by a comment id. A graphql resolve that names no `PRRT_` id is denied. A reply is an `add_reply_to_pull_request_comment` call, or a `gh api` call to a comment's `replies` endpoint or to `addPullRequestReviewThreadReply`, which names its thread by `PRRT_` id. A reply and a resolve by comment id match when they name the same comment, or when a successful tool result in the transcript, such as `pull_request_read` `get_review_comments`, lists the comment under the thread by its `discussion_r` anchor or `databaseId`. A reply whose result is an error, such as a refusal from `blocking/reply_length_gate.py`, does not count. With no readable transcript, the resolve passes.
|
|
22
22
|
- `blocking/verify_before_commit_gate.py` denies a `git commit` in a repository with a `.claude/skills/verify` or `.agents/skills/verify` directory until the `verify` skill loads after the last user prompt, through a Skill call whose result comes back without an error, or the `/verify` command in that prompt. It reads the committed paths from the staged diff, the tracked changes for `-a`, and the paths of each `git add` earlier in the same command, and passes a commit whose paths are all docs, images, test files, or `.json`, `.jsonl`, `.yaml`, `.yml`, or `.csv` files under a `docs` folder.
|
|
23
23
|
- `blocking/build_eval_before_eval_write_gate.py` denies a write to an eval file until a Skill call named exactly `claude-api` with args `build-eval`, or the `/claude-api build-eval` command, appears in the transcript.
|
|
@@ -59,13 +59,43 @@ GIT_ACTIONS = frozenset({"commit", "push"})
|
|
|
59
59
|
GH_PULL_REQUEST_COMMAND = "pr"
|
|
60
60
|
GH_API_COMMAND = "api"
|
|
61
61
|
MERGE_PATH_SUFFIX = "/merge"
|
|
62
|
-
ALL_API_ACTION_NAMES = ("auto-merge", "enablePullRequestAutoMerge")
|
|
62
|
+
ALL_API_ACTION_NAMES = ("auto-merge", "enablePullRequestAutoMerge", "mergePullRequest", "createPullRequest")
|
|
63
|
+
COMMAND_SUBSTITUTION_PATTERN = re.compile(r"\$\(([^()]*)\)|`([^`]*)`")
|
|
64
|
+
DOLLAR_SUBSTITUTION_GROUP = 1
|
|
65
|
+
BACKTICK_SUBSTITUTION_GROUP = 2
|
|
63
66
|
ALL_COMMAND_PREFIX_WORDS = frozenset({"env", "sudo", "command", "exec", "nohup", "time"})
|
|
67
|
+
ALL_GH_PULL_REQUEST_READ_SUBCOMMANDS = frozenset({"checks", "diff", "list", "status", "view"})
|
|
68
|
+
ALL_GH_PULL_REQUEST_SUBCOMMANDS = ALL_GH_PULL_REQUEST_READ_SUBCOMMANDS | frozenset(
|
|
69
|
+
{
|
|
70
|
+
"checkout",
|
|
71
|
+
"close",
|
|
72
|
+
"comment",
|
|
73
|
+
"create",
|
|
74
|
+
"edit",
|
|
75
|
+
"lock",
|
|
76
|
+
"merge",
|
|
77
|
+
"ready",
|
|
78
|
+
"reopen",
|
|
79
|
+
"review",
|
|
80
|
+
"revert",
|
|
81
|
+
"unlock",
|
|
82
|
+
"update-branch",
|
|
83
|
+
}
|
|
84
|
+
)
|
|
64
85
|
ALL_PREFIX_OPTIONS_WITH_VALUE = frozenset({"-u", "--unset", "-C", "--chdir", "-g", "--group", "--user"})
|
|
65
86
|
ENVIRONMENT_ASSIGNMENT_PATTERN = r"^[A-Za-z_][A-Za-z0-9_]*="
|
|
66
|
-
ALL_SHELL_WRAPPER_EXECUTABLES = frozenset({"bash", "sh", "zsh", "dash", "pwsh", "pwsh.exe", "powershell", "powershell.exe"})
|
|
87
|
+
ALL_SHELL_WRAPPER_EXECUTABLES = frozenset({"bash", "sh", "zsh", "dash", "eval", "pwsh", "pwsh.exe", "powershell", "powershell.exe"})
|
|
67
88
|
ALL_PYTHON_EXECUTABLES = frozenset({"python", "python3", "python.exe", "python3.exe", "py", "py.exe"})
|
|
68
89
|
PULL_REQUEST_SCRIPT_NAME = "pull_request.py"
|
|
90
|
+
EVAL_PROGRAM_NAME = "eval"
|
|
91
|
+
EVAL_ARGUMENT_SEPARATOR = " "
|
|
92
|
+
ALL_NESTED_PROGRAM_WRAPPER_NAMES = frozenset({"xargs", "timeout", "nice", "ionice", "stdbuf", "parallel", "watch", "find"})
|
|
93
|
+
ALL_NESTED_PROGRAM_NAMES = (
|
|
94
|
+
frozenset({"gh", "gh.exe", "git", "git.exe", PULL_REQUEST_SCRIPT_NAME})
|
|
95
|
+
| ALL_PYTHON_EXECUTABLES
|
|
96
|
+
| ALL_SHELL_WRAPPER_EXECUTABLES
|
|
97
|
+
| ALL_NESTED_PROGRAM_WRAPPER_NAMES
|
|
98
|
+
)
|
|
69
99
|
PATH_SEPARATOR_PATTERN = r"[\\/]"
|
|
70
100
|
PYTHON_OPTIONS_WITH_VALUE = frozenset({"-X", "-W"})
|
|
71
101
|
ALL_BUILD_EVAL_SKILL_NAMES = ("claude-api",)
|
|
@@ -2,8 +2,8 @@
|
|
|
2
2
|
|
|
3
3
|
import re
|
|
4
4
|
|
|
5
|
-
MAXIMUM_SENTENCE_COUNT =
|
|
6
|
-
MAXIMUM_WORDS_PER_SENTENCE =
|
|
5
|
+
MAXIMUM_SENTENCE_COUNT = 50
|
|
6
|
+
MAXIMUM_WORDS_PER_SENTENCE = 35
|
|
7
7
|
ALL_CHECKED_TOOL_NAMES = frozenset(
|
|
8
8
|
{"mcp__hearthbot__reply", "mcp__hearthbot__post_message", "mcp__hearthbot__update_message"}
|
|
9
9
|
)
|
|
@@ -182,6 +182,6 @@ ALL_GH_BODY_FLAGS = frozenset({"--body", "-b"})
|
|
|
182
182
|
GH_BODY_FLAG_PREFIX = "--body="
|
|
183
183
|
DETAILS_BLOCK_PATTERN = re.compile(r"<details\b.*?</details>", re.IGNORECASE | re.DOTALL)
|
|
184
184
|
COMMENT_RETRY_INSTRUCTION = (
|
|
185
|
-
" A comment on GitHub holds at most
|
|
185
|
+
f" A comment on GitHub holds at most {MAXIMUM_SENTENCE_COUNT} short sentences."
|
|
186
186
|
" Put detail in the pull request body, a code block, or a collapsed <details> block, then post again."
|
|
187
187
|
)
|
|
@@ -1,8 +1,8 @@
|
|
|
1
1
|
from hooks_constants import reply_length_gate_constants as constants
|
|
2
2
|
|
|
3
3
|
|
|
4
|
-
def
|
|
5
|
-
assert (constants.MAXIMUM_SENTENCE_COUNT, constants.MAXIMUM_WORDS_PER_SENTENCE) == (
|
|
4
|
+
def test_gate_limits_replies_to_fifty_sentences_of_thirty_five_words() -> None:
|
|
5
|
+
assert (constants.MAXIMUM_SENTENCE_COUNT, constants.MAXIMUM_WORDS_PER_SENTENCE) == (50, 35)
|
|
6
6
|
|
|
7
7
|
|
|
8
8
|
def test_gate_checks_the_three_chat_tools_that_post_to_the_user() -> None:
|
package/package.json
CHANGED
|
@@ -10,6 +10,10 @@ from pathlib import Path
|
|
|
10
10
|
import pytest
|
|
11
11
|
|
|
12
12
|
import durable_post_lint
|
|
13
|
+
from hooks_constants.reply_length_gate_constants import (
|
|
14
|
+
MAXIMUM_SENTENCE_COUNT,
|
|
15
|
+
MAXIMUM_WORDS_PER_SENTENCE,
|
|
16
|
+
)
|
|
13
17
|
from dev_env_scripts_constants.private_term_constants import PrivateTermDigest
|
|
14
18
|
from hooks_constants.pull_request_proof_constants import (
|
|
15
19
|
MISSING_EVAL_SECTION_REASON,
|
|
@@ -62,9 +66,8 @@ Ran `python scripts/durable_post_lint.py --action pr-create --body-file body.md`
|
|
|
62
66
|
"""
|
|
63
67
|
OVER_LONG_REPLY_BODY = (
|
|
64
68
|
"Fixed in `abc1234`.\n\n"
|
|
65
|
-
"
|
|
66
|
-
"
|
|
67
|
-
"---\nPosted by the review agent."
|
|
69
|
+
+ " ".join(f"Status line {each_number} is ready." for each_number in range(MAXIMUM_SENTENCE_COUNT + 1))
|
|
70
|
+
+ "\n\n---\nPosted by the review agent."
|
|
68
71
|
)
|
|
69
72
|
MINIMAL_PR_BODY = """## Why
|
|
70
73
|
|
|
@@ -444,7 +447,7 @@ def should_exit_with_findings_for_a_create_body_with_no_existing_work_section(
|
|
|
444
447
|
assert completed_process.stderr == f"{MISSING_EXISTING_WORK_REASON}\n"
|
|
445
448
|
|
|
446
449
|
|
|
447
|
-
def
|
|
450
|
+
def should_report_a_review_reply_over_the_sentence_limit() -> None:
|
|
448
451
|
all_findings = durable_post_lint.lint_durable_post(
|
|
449
452
|
action="github-mcp-comment",
|
|
450
453
|
title=None,
|
|
@@ -455,13 +458,12 @@ def should_report_a_review_reply_over_three_sentences() -> None:
|
|
|
455
458
|
]
|
|
456
459
|
|
|
457
460
|
|
|
458
|
-
def
|
|
461
|
+
def should_report_a_review_reply_sentence_over_the_word_limit() -> None:
|
|
459
462
|
all_findings = durable_post_lint.lint_durable_post(
|
|
460
463
|
action="github-mcp-comment",
|
|
461
464
|
title=None,
|
|
462
465
|
body_text=(
|
|
463
|
-
"
|
|
464
|
-
" caller on the same import path."
|
|
466
|
+
" ".join(f"word{each_number}" for each_number in range(MAXIMUM_WORDS_PER_SENTENCE + 1)) + "."
|
|
465
467
|
),
|
|
466
468
|
)
|
|
467
469
|
assert [each_finding.code for each_finding in all_findings] == [
|
|
@@ -495,7 +497,7 @@ def should_exit_with_findings_for_an_over_long_review_reply(tmp_path: Path) -> N
|
|
|
495
497
|
text=True,
|
|
496
498
|
)
|
|
497
499
|
assert completed_process.returncode == 1
|
|
498
|
-
assert "at most
|
|
500
|
+
assert f"at most {MAXIMUM_SENTENCE_COUNT} short sentences" in completed_process.stderr
|
|
499
501
|
|
|
500
502
|
|
|
501
503
|
def should_exit_clean_for_a_short_review_reply(tmp_path: Path) -> None:
|