claude-dev-env 8.117.3 → 8.117.4
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/_shared/pr-loop/scripts/pyproject.toml +1 -1
- package/_shared/process-tree/scripts/pyproject.toml +1 -1
- package/bin/cde.mjs +4 -3
- package/bin/cde.test.mjs +87 -1
- package/bin/codex-compat.mjs +4 -3
- package/bin/codex-compat.test.mjs +91 -1
- package/bin/install-plan.mjs +2 -1
- package/bin/install.mjs +24 -7
- package/bin/install.plan.test.mjs +15 -0
- package/bin/install.test.mjs +61 -0
- package/bin/python-probe.mjs +42 -0
- package/bin/python-probe.test.mjs +19 -2
- package/docs/rule-guides/features-start-with-an-eval.md +3 -3
- package/docs/rule-guides/proof-before-pull-request.md +2 -2
- package/hooks/blocking/bash_pre_tool_use_dispatcher.py +11 -4
- package/hooks/blocking/build_eval_before_eval_write_gate.py +10 -4
- package/hooks/blocking/followup_pr_dedupe.py +381 -64
- package/hooks/blocking/gate_child_process.py +187 -0
- package/hooks/blocking/gate_run_reads.py +116 -35
- package/hooks/blocking/merge_review_gate.py +93 -10
- package/hooks/blocking/merge_review_gate_parts/create_folders.py +346 -0
- package/hooks/blocking/merge_review_gate_parts/create_options.py +86 -0
- package/hooks/blocking/merge_review_gate_parts/folder_state.py +107 -0
- package/hooks/blocking/merge_review_gate_parts/open_request.py +199 -64
- package/hooks/blocking/merge_review_gate_parts/quoted_lines.py +283 -0
- package/hooks/blocking/merge_review_gate_parts/wrapped_creates.py +126 -0
- package/hooks/blocking/pr_lifecycle_skill_gate.py +61 -18
- package/hooks/blocking/pre_tool_use_dispatcher.py +25 -3
- package/hooks/blocking/pull_request_proof.py +5 -51
- package/hooks/blocking/pull_request_proof_edit_gate.py +19 -5
- package/hooks/blocking/test_bash_pre_tool_use_dispatcher.py +54 -0
- package/hooks/blocking/test_build_eval_before_eval_write_gate.py +64 -0
- package/hooks/blocking/test_followup_pr_dedupe.py +464 -1
- package/hooks/blocking/test_gate_child_process.py +64 -0
- package/hooks/blocking/test_gate_run_reads.py +364 -4
- package/hooks/blocking/test_merge_review_gate.py +832 -39
- package/hooks/blocking/test_pr_lifecycle_skill_gate.py +130 -4
- package/hooks/blocking/test_pre_tool_use_dispatcher.py +64 -0
- package/hooks/blocking/test_pull_request_proof.py +16 -0
- package/hooks/blocking/test_pull_request_proof_edit_gate.py +60 -0
- package/hooks/blocking/tests/test_create_folders.py +290 -0
- package/hooks/blocking/tests/test_create_options.py +34 -0
- package/hooks/blocking/tests/test_folder_state.py +40 -0
- package/hooks/blocking/tests/test_open_request.py +350 -44
- package/hooks/blocking/tests/test_quoted_lines.py +86 -0
- package/hooks/blocking/tests/test_wrapped_creates.py +42 -0
- package/hooks/features/bash-dispatchers.md +1 -1
- package/hooks/features/conduct-gates.md +7 -7
- package/hooks/features/write-edit-blocking.md +3 -3
- package/hooks/hooks_constants/build_eval_before_eval_write_gate_constants.py +5 -0
- package/hooks/hooks_constants/followup_pr_dedupe_constants.py +34 -1
- package/hooks/hooks_constants/gate_fault.py +52 -0
- package/hooks/hooks_constants/gate_fault_constants.py +13 -0
- package/hooks/hooks_constants/gate_run_reads_constants.py +9 -1
- package/hooks/hooks_constants/merge_review_gate_constants.py +94 -1
- package/hooks/hooks_constants/piped_pytest_blocker_constants.py +2 -0
- package/hooks/hooks_constants/pr_lifecycle_skill_gate_constants.py +6 -1
- package/hooks/hooks_constants/pre_tool_use_stdin.py +23 -3
- package/hooks/hooks_constants/shell_command_pipeline.py +68 -8
- package/hooks/hooks_constants/shell_command_segments.py +2 -2
- package/hooks/hooks_constants/shell_command_wrappers.py +3 -2
- package/hooks/hooks_constants/test_followup_pr_dedupe_constants.py +7 -0
- package/hooks/hooks_constants/test_gate_fault.py +41 -0
- package/hooks/hooks_constants/test_piped_pytest_blocker_constants.py +5 -0
- package/hooks/hooks_constants/test_pr_lifecycle_skill_gate_constants.py +10 -0
- package/hooks/hooks_constants/test_pre_tool_use_stdin.py +97 -0
- package/hooks/hooks_constants/test_shell_command_pipeline.py +55 -0
- package/hooks/hooks_constants/test_shell_command_segments.py +15 -0
- package/hooks/hooks_constants/test_shell_command_wrappers.py +6 -0
- package/hooks/hooks_constants/transcript_skill_scan_constants.py +6 -0
- package/hooks/pyproject.toml +2 -3
- package/hooks/test_python_floor.py +197 -0
- package/hooks/test_transcript_skill_scan.py +324 -0
- package/hooks/transcript_skill_scan.py +233 -2
- package/package.json +1 -1
|
@@ -0,0 +1,126 @@
|
|
|
1
|
+
"""Find each create that a wrapped inner text holds and the folder walk did not place.
|
|
2
|
+
|
|
3
|
+
::
|
|
4
|
+
|
|
5
|
+
bash -c "cd sub && gh pr create --fill" -> ("--fill",) when the walk missed it
|
|
6
|
+
gh pr create --body "a\\ngh pr create" -> nothing: a quoted value runs no command
|
|
7
|
+
|
|
8
|
+
``hooks_constants.shell_command_wrappers.all_wrapped_command_texts`` lists the
|
|
9
|
+
command and each text a wrapper or substitution inside it runs. A text is
|
|
10
|
+
skipped when it is the command itself, or a closed text with its line breaks
|
|
11
|
+
read as spaces, or a text unwrapped from one. Each other text is cut into POSIX
|
|
12
|
+
words, and each create it holds is compared with the creates the walk placed,
|
|
13
|
+
line breaks read as spaces.
|
|
14
|
+
"""
|
|
15
|
+
|
|
16
|
+
from __future__ import annotations
|
|
17
|
+
|
|
18
|
+
import sys
|
|
19
|
+
from collections.abc import Sequence
|
|
20
|
+
from pathlib import Path
|
|
21
|
+
|
|
22
|
+
hooks_directory = str(Path(__file__).resolve().parent.parent.parent)
|
|
23
|
+
if hooks_directory not in sys.path:
|
|
24
|
+
sys.path.insert(0, hooks_directory)
|
|
25
|
+
|
|
26
|
+
from blocking.followup_pr_dedupe import create_arguments
|
|
27
|
+
from blocking.merge_review_gate_parts.quoted_lines import quote_aware_lines
|
|
28
|
+
from hooks_constants.merge_review_gate_constants import (
|
|
29
|
+
IS_POSIX_TOKENIZATION,
|
|
30
|
+
SUBSHELL_CLOSE_TOKEN,
|
|
31
|
+
)
|
|
32
|
+
from hooks_constants.pytest_invocation import unquoted_token
|
|
33
|
+
from hooks_constants.shell_command_evaluation_constants import (
|
|
34
|
+
ALL_LINE_BREAK_CHARACTERS,
|
|
35
|
+
QUOTED_LINE_BREAK_REPLACEMENT,
|
|
36
|
+
)
|
|
37
|
+
from hooks_constants.shell_command_pipeline import (
|
|
38
|
+
join_line_continuations,
|
|
39
|
+
operator_aware_tokens,
|
|
40
|
+
scannable_command_lines,
|
|
41
|
+
segments_with_following_operator,
|
|
42
|
+
)
|
|
43
|
+
from hooks_constants.shell_command_wrappers import (
|
|
44
|
+
all_wrapped_command_texts,
|
|
45
|
+
quoted_line_breaks_joined,
|
|
46
|
+
)
|
|
47
|
+
|
|
48
|
+
|
|
49
|
+
def unplaced_creates(
|
|
50
|
+
command: str, all_placed_arguments: Sequence[Sequence[str]]
|
|
51
|
+
) -> list[tuple[tuple[str, ...], tuple[str, ...]]]:
|
|
52
|
+
"""Return the arguments and command words of each wrapped create the walk did not place.
|
|
53
|
+
|
|
54
|
+
::
|
|
55
|
+
|
|
56
|
+
'bash -c "gh pr create --fill"', [] -> [(("--fill",), ("gh", "pr", "create", "--fill"))]
|
|
57
|
+
'bash -c "gh pr create --fill"', [("--fill",)] -> []
|
|
58
|
+
|
|
59
|
+
Args:
|
|
60
|
+
command: The shell command text.
|
|
61
|
+
all_placed_arguments: The arguments of each create the walk placed.
|
|
62
|
+
|
|
63
|
+
Returns:
|
|
64
|
+
Each unplaced create's arguments and the words of the simple command that runs it,
|
|
65
|
+
in the order the wrapped texts list them.
|
|
66
|
+
"""
|
|
67
|
+
all_known = {_normalized(each_arguments) for each_arguments in all_placed_arguments}
|
|
68
|
+
return [
|
|
69
|
+
(tuple(all_arguments), tuple(all_words))
|
|
70
|
+
for all_arguments, all_words in _wrapped_creates(command)
|
|
71
|
+
if _normalized(all_arguments) not in all_known
|
|
72
|
+
]
|
|
73
|
+
|
|
74
|
+
|
|
75
|
+
def _normalized(all_arguments: Sequence[str]) -> tuple[str, ...]:
|
|
76
|
+
all_words = [
|
|
77
|
+
_line_breaks_as_spaces(unquoted_token(each_argument)) for each_argument in all_arguments
|
|
78
|
+
]
|
|
79
|
+
while all_words and all_words[-1] == SUBSHELL_CLOSE_TOKEN:
|
|
80
|
+
all_words.pop()
|
|
81
|
+
return tuple(all_words)
|
|
82
|
+
|
|
83
|
+
|
|
84
|
+
def _line_breaks_as_spaces(word: str) -> str:
|
|
85
|
+
return "".join(
|
|
86
|
+
QUOTED_LINE_BREAK_REPLACEMENT
|
|
87
|
+
if each_character in ALL_LINE_BREAK_CHARACTERS
|
|
88
|
+
else each_character
|
|
89
|
+
for each_character in word
|
|
90
|
+
)
|
|
91
|
+
|
|
92
|
+
|
|
93
|
+
def _wrapped_creates(command: str) -> list[tuple[list[str], list[str]]]:
|
|
94
|
+
all_wrapped_texts = all_wrapped_command_texts(command)
|
|
95
|
+
all_skipped_texts = {
|
|
96
|
+
command,
|
|
97
|
+
quoted_line_breaks_joined(command),
|
|
98
|
+
*_flattened_balanced_texts(all_wrapped_texts),
|
|
99
|
+
}
|
|
100
|
+
return [
|
|
101
|
+
(each_arguments, each_segment)
|
|
102
|
+
for each_text in all_wrapped_texts
|
|
103
|
+
if each_text not in all_skipped_texts
|
|
104
|
+
for each_segment in _posix_segments(each_text)
|
|
105
|
+
if (each_arguments := create_arguments(each_segment)) is not None
|
|
106
|
+
]
|
|
107
|
+
|
|
108
|
+
|
|
109
|
+
def _flattened_balanced_texts(all_texts: list[str]) -> set[str]:
|
|
110
|
+
return {
|
|
111
|
+
each_unwrapped
|
|
112
|
+
for each_text in all_texts
|
|
113
|
+
if quote_aware_lines(each_text) is not None
|
|
114
|
+
and (each_flattened := _line_breaks_as_spaces(each_text)) != each_text
|
|
115
|
+
for each_unwrapped in all_wrapped_command_texts(each_flattened)
|
|
116
|
+
}
|
|
117
|
+
|
|
118
|
+
|
|
119
|
+
def _posix_segments(text: str) -> list[list[str]]:
|
|
120
|
+
return [
|
|
121
|
+
each_segment
|
|
122
|
+
for each_line in scannable_command_lines(join_line_continuations(text))
|
|
123
|
+
for each_segment, _each_operator in segments_with_following_operator(
|
|
124
|
+
operator_aware_tokens(each_line, IS_POSIX_TOKENIZATION) or []
|
|
125
|
+
)
|
|
126
|
+
]
|
|
@@ -1,7 +1,10 @@
|
|
|
1
1
|
"""Deny pull request lifecycle actions until their skill appears in the transcript.
|
|
2
2
|
|
|
3
|
-
A gate silent at its 30 s ``hooks.json`` timeout lets the call through, so its git and GitHub reads
|
|
4
|
-
``blocking/gate_run_reads.py`` draws that budget.
|
|
3
|
+
A gate silent at its 30 s ``hooks.json`` timeout lets the call through, so its git and GitHub reads
|
|
4
|
+
and its transcript scans end by 21 s. ``blocking/gate_run_reads.py`` draws that budget.
|
|
5
|
+
A deny from a run that cut a read short says a second run may pass.
|
|
6
|
+
Each transcript scan reads the newest line first, so a recent skill load answers early.
|
|
7
|
+
A deny from a scan the deadline stopped says to load the skill, or call build-eval, again.
|
|
5
8
|
"""
|
|
6
9
|
|
|
7
10
|
from __future__ import annotations
|
|
@@ -44,6 +47,7 @@ from hooks_constants.pr_lifecycle_skill_gate_constants import (
|
|
|
44
47
|
DOUBLE_QUOTED_ESCAPE_PATTERN,
|
|
45
48
|
ENVIRONMENT_ASSIGNMENT_PATTERN,
|
|
46
49
|
ESCAPED_CHARACTER_GROUP,
|
|
50
|
+
GATE_NAME,
|
|
47
51
|
GH_API_COMMAND,
|
|
48
52
|
GH_PULL_REQUEST_COMMAND,
|
|
49
53
|
GIT_ACTIONS,
|
|
@@ -57,6 +61,7 @@ from hooks_constants.pr_lifecycle_skill_gate_constants import (
|
|
|
57
61
|
PATH_SEPARATOR_PATTERN,
|
|
58
62
|
PERMISSION_DECISION_REASON_KEY,
|
|
59
63
|
PULL_REQUEST_SCRIPT_NAME,
|
|
64
|
+
SCAN_RAN_OUT_SENTENCE,
|
|
60
65
|
ALL_NESTED_PROGRAM_NAMES,
|
|
61
66
|
ALL_NESTED_PROGRAM_WRAPPER_NAMES,
|
|
62
67
|
EVAL_ARGUMENT_SEPARATOR,
|
|
@@ -78,14 +83,21 @@ from hooks_constants.hook_specific_output_keys import (
|
|
|
78
83
|
)
|
|
79
84
|
from hooks_constants.pre_tool_use_stdin import print_decision_from_stdin
|
|
80
85
|
from hooks_constants.setup_project_paths_constants import DECODE_ERRORS_POLICY, UTF8_ENCODING
|
|
81
|
-
from blocking.followup_pr_dedupe import create_tool_fields, duplicate_followup_reason
|
|
82
|
-
from blocking.gate_run_reads import
|
|
86
|
+
from blocking.followup_pr_dedupe import all_create_argument_lists, create_tool_fields, duplicate_followup_reason
|
|
87
|
+
from blocking.gate_run_reads import (
|
|
88
|
+
GateRun,
|
|
89
|
+
ReadSlice,
|
|
90
|
+
gate_run,
|
|
91
|
+
read_deadlines_started,
|
|
92
|
+
transcript_lines_within_slice,
|
|
93
|
+
)
|
|
94
|
+
from hooks_constants.gate_fault import gate_fault_reason
|
|
83
95
|
from hooks_constants.gate_run_reads_constants import READS_RAN_OUT_SENTENCE
|
|
84
96
|
from blocking import pull_request_proof as proof
|
|
85
97
|
from hooks_constants.pull_request_proof_constants import MISSING_BUILD_EVAL_REASON
|
|
86
98
|
from transcript_skill_scan import (
|
|
87
99
|
invokes_skill_with_argument,
|
|
88
|
-
|
|
100
|
+
is_skill_loaded_after_last_compaction_newest_first,
|
|
89
101
|
)
|
|
90
102
|
|
|
91
103
|
|
|
@@ -302,6 +314,8 @@ def _wrapped_word_lists(all_command_words: list[str]) -> list[list[str]]:
|
|
|
302
314
|
|
|
303
315
|
|
|
304
316
|
def _matches_shell_command(command: str) -> bool:
|
|
317
|
+
if next(all_create_argument_lists(command), None) is not None:
|
|
318
|
+
return True
|
|
305
319
|
for each_command_words in executable_word_lists(command):
|
|
306
320
|
executable = file_name(each_command_words[0])
|
|
307
321
|
if executable == PULL_REQUEST_SCRIPT_NAME:
|
|
@@ -327,17 +341,17 @@ def _is_governed_action(all_payload_fields: dict[str, object]) -> bool:
|
|
|
327
341
|
return isinstance(tool_input, dict) and isinstance(tool_input.get("command"), str) and _matches_shell_command(tool_input["command"])
|
|
328
342
|
|
|
329
343
|
|
|
330
|
-
def _scanned_transcript(path: str, scan: Callable[[Iterable[str]], bool]) -> bool | None:
|
|
344
|
+
def _scanned_transcript(path: str, scan: Callable[[Iterable[str]], bool], read_slice: ReadSlice) -> bool | None:
|
|
331
345
|
try:
|
|
332
|
-
with open(path,
|
|
333
|
-
return scan(transcript)
|
|
346
|
+
with open(path, "rb") as transcript:
|
|
347
|
+
return scan(transcript_lines_within_slice(transcript, UTF8_ENCODING, DECODE_ERRORS_POLICY, read_slice))
|
|
334
348
|
except OSError:
|
|
335
349
|
return None
|
|
336
350
|
|
|
337
351
|
|
|
338
|
-
def _shows_skill_load(
|
|
339
|
-
return
|
|
340
|
-
|
|
352
|
+
def _shows_skill_load(all_lines_newest_first: Iterable[str]) -> bool:
|
|
353
|
+
return is_skill_loaded_after_last_compaction_newest_first(
|
|
354
|
+
all_lines_newest_first, (SKILL_NAME,), ALL_SLASH_COMMAND_MARKERS
|
|
341
355
|
)
|
|
342
356
|
|
|
343
357
|
|
|
@@ -381,9 +395,14 @@ def decision_for(all_payload_fields: dict[str, object]) -> dict[str, object] | N
|
|
|
381
395
|
changed page whose proof shows no picture -> MISSING_LOOK_REASON
|
|
382
396
|
feature pull request, no /claude-api build-eval -> MISSING_BUILD_EVAL_REASON
|
|
383
397
|
second follow-up pull request for one parent -> DUPLICATE_FOLLOWUP_REASON_TEMPLATE
|
|
398
|
+
follow-up read of open pull requests unanswered -> UNANSWERED_FOLLOWUP_REASON_TEMPLATE
|
|
399
|
+
follow-up with no repository, origin unread -> UNREAD_ORIGIN_FOLLOWUP_REASON_TEMPLATE
|
|
400
|
+
skill or build-eval deny from a scan the deadline stopped -> that reason + SCAN_RAN_OUT_SENTENCE
|
|
384
401
|
any deny above in a run whose deadline cut a read -> that reason + READS_RAN_OUT_SENTENCE
|
|
402
|
+
a check raises once the call is known governed -> gate_fault_reason naming the error
|
|
385
403
|
|
|
386
404
|
The checks share one gate run, so two checks that ask git or GitHub one question read it once.
|
|
405
|
+
A fault while the gate decides whether the call is governed still raises, so it never refuses an unrelated call.
|
|
387
406
|
|
|
388
407
|
Args:
|
|
389
408
|
all_payload_fields: The parsed PreToolUse input.
|
|
@@ -391,14 +410,31 @@ def decision_for(all_payload_fields: dict[str, object]) -> dict[str, object] | N
|
|
|
391
410
|
if not _is_governed_action(all_payload_fields):
|
|
392
411
|
return None
|
|
393
412
|
with gate_run() as run:
|
|
394
|
-
|
|
413
|
+
try:
|
|
414
|
+
reason = _first_broken_rule_reason(all_payload_fields, run)
|
|
415
|
+
except Exception as error:
|
|
416
|
+
return deny_decision(gate_fault_reason(GATE_NAME, error))
|
|
395
417
|
note = READS_RAN_OUT_SENTENCE if run.was_read_cut_short else ""
|
|
396
418
|
return None if reason is None else deny_decision(reason + note)
|
|
397
419
|
|
|
398
420
|
|
|
399
|
-
def
|
|
421
|
+
def _scan_note(run: GateRun, cut_scan_count_before: int) -> str:
|
|
422
|
+
return SCAN_RAN_OUT_SENTENCE if run.cut_scan_count > cut_scan_count_before else ""
|
|
423
|
+
|
|
424
|
+
|
|
425
|
+
def _missing_build_eval_reason(all_fields: dict[str, object], run: GateRun) -> str | None:
|
|
426
|
+
if not proof.is_feature_pull_request(all_fields):
|
|
427
|
+
return None
|
|
428
|
+
cut_scan_count_before = run.cut_scan_count
|
|
429
|
+
if not is_build_eval_missing(all_fields):
|
|
430
|
+
return None
|
|
431
|
+
return MISSING_BUILD_EVAL_REASON + _scan_note(run, cut_scan_count_before)
|
|
432
|
+
|
|
433
|
+
|
|
434
|
+
def _first_broken_rule_reason(all_fields: dict[str, object], run: GateRun) -> str | None:
|
|
435
|
+
cut_scan_count_before = run.cut_scan_count
|
|
400
436
|
if _is_skill_unloaded(all_fields):
|
|
401
|
-
return DENY_REASON
|
|
437
|
+
return DENY_REASON + _scan_note(run, cut_scan_count_before)
|
|
402
438
|
return (
|
|
403
439
|
proof.missing_proof_reason(all_fields)
|
|
404
440
|
or proof.edit_proof_reason(all_fields)
|
|
@@ -406,7 +442,7 @@ def _first_broken_rule_reason(all_fields: dict[str, object]) -> str | None:
|
|
|
406
442
|
or proof.missing_flowchart_reason(all_fields)
|
|
407
443
|
or proof.missing_eval_section_reason(all_fields)
|
|
408
444
|
or proof.missing_look_reason(all_fields)
|
|
409
|
-
or (
|
|
445
|
+
or _missing_build_eval_reason(all_fields, run)
|
|
410
446
|
or duplicate_followup_reason(all_fields)
|
|
411
447
|
)
|
|
412
448
|
|
|
@@ -431,11 +467,13 @@ def _is_skill_unloaded(all_payload_fields: dict[str, object]) -> bool:
|
|
|
431
467
|
readable_paths = transcript_paths(all_payload_fields)
|
|
432
468
|
if not readable_paths:
|
|
433
469
|
return False
|
|
434
|
-
statuses = [_scanned_transcript(path, _shows_skill_load) for path in readable_paths]
|
|
470
|
+
statuses = [_scanned_transcript(path, _shows_skill_load, ReadSlice.PROOF_AND_LOOK) for path in readable_paths]
|
|
435
471
|
return not any(status is None or status for status in statuses)
|
|
436
472
|
|
|
437
473
|
|
|
438
|
-
def eval_skill_call_status(
|
|
474
|
+
def eval_skill_call_status(
|
|
475
|
+
all_payload_fields: dict[str, object], read_slice: ReadSlice = ReadSlice.FOLLOWUP
|
|
476
|
+
) -> bool | None:
|
|
439
477
|
"""Return whether the session's transcripts show a /claude-api build-eval call.
|
|
440
478
|
|
|
441
479
|
::
|
|
@@ -443,13 +481,18 @@ def eval_skill_call_status(all_payload_fields: dict[str, object]) -> bool | None
|
|
|
443
481
|
one transcript shows the call -> True
|
|
444
482
|
transcripts read, none shows the call -> False
|
|
445
483
|
no transcript could be read -> None
|
|
484
|
+
the slice's deadline stopped a scan -> False, unless another transcript shows the call
|
|
485
|
+
|
|
486
|
+
Each scan reads the newest line first and stops at the slice's deadline
|
|
487
|
+
once the gate process started its read deadlines.
|
|
446
488
|
|
|
447
489
|
Args:
|
|
448
490
|
all_payload_fields: The parsed PreToolUse input.
|
|
491
|
+
read_slice: The slice whose deadline each scan answers to.
|
|
449
492
|
"""
|
|
450
493
|
has_read_transcript = False
|
|
451
494
|
for each_path in transcript_paths(all_payload_fields):
|
|
452
|
-
status = _scanned_transcript(each_path, _shows_build_eval)
|
|
495
|
+
status = _scanned_transcript(each_path, _shows_build_eval, read_slice)
|
|
453
496
|
if status:
|
|
454
497
|
return True
|
|
455
498
|
has_read_transcript = has_read_transcript or status is not None
|
|
@@ -510,6 +510,22 @@ def _emit_allow_decision(decision: DispatcherDecision) -> None:
|
|
|
510
510
|
sys.stdout.flush()
|
|
511
511
|
|
|
512
512
|
|
|
513
|
+
def _emit_system_message_only(decision: DispatcherDecision) -> None:
|
|
514
|
+
"""Write one JSON object that carries only the hooks' systemMessage lines.
|
|
515
|
+
|
|
516
|
+
A hosted gate whose check raised prints a warning and no decision. This
|
|
517
|
+
payload shows that warning to the user and carries no permissionDecision,
|
|
518
|
+
so the call goes on to the normal permission flow.
|
|
519
|
+
|
|
520
|
+
Args:
|
|
521
|
+
decision: The aggregated dispatcher decision with no deny and no allow.
|
|
522
|
+
"""
|
|
523
|
+
all_unique_system_messages = unique_first_seen_strings(decision.all_system_messages)
|
|
524
|
+
message_payload = {"systemMessage": SYSTEM_MESSAGE_JOIN_SEPARATOR.join(all_unique_system_messages)}
|
|
525
|
+
sys.stdout.write(json.dumps(message_payload) + "\n")
|
|
526
|
+
sys.stdout.flush()
|
|
527
|
+
|
|
528
|
+
|
|
513
529
|
def _select_applicable_hooks(tool_name: str) -> list[HostedHookEntry]:
|
|
514
530
|
"""Return the ordered hosted hook entries applicable to the given tool name.
|
|
515
531
|
|
|
@@ -573,9 +589,12 @@ def dispatch(
|
|
|
573
589
|
|
|
574
590
|
Selects the applicable hosted hooks for tool_name, runs each one in-process
|
|
575
591
|
(natively when the entry names a native module, otherwise via runpy),
|
|
576
|
-
aggregates the results, and emits
|
|
577
|
-
|
|
578
|
-
|
|
592
|
+
aggregates the results, and emits one of four outcomes::
|
|
593
|
+
|
|
594
|
+
any hook denied -> deny JSON object
|
|
595
|
+
a hook allowed explicitly and none denied -> allow JSON object
|
|
596
|
+
no hook decided, a hook printed systemMessage -> {"systemMessage": ...} alone
|
|
597
|
+
no hook decided or spoke -> no output
|
|
579
598
|
|
|
580
599
|
Args:
|
|
581
600
|
payload_text: The raw JSON payload text to replay to each runpy hook.
|
|
@@ -594,6 +613,9 @@ def dispatch(
|
|
|
594
613
|
return
|
|
595
614
|
if aggregated_decision.should_allow:
|
|
596
615
|
_emit_allow_decision(aggregated_decision)
|
|
616
|
+
return
|
|
617
|
+
if aggregated_decision.all_system_messages:
|
|
618
|
+
_emit_system_message_only(aggregated_decision)
|
|
597
619
|
|
|
598
620
|
|
|
599
621
|
def main() -> None:
|
|
@@ -15,7 +15,8 @@ if hooks_directory not in sys.path:
|
|
|
15
15
|
|
|
16
16
|
from blocking.body_flowchart import fenced_line_flags, section_has_flowchart
|
|
17
17
|
from blocking.followup_pr_dedupe import (
|
|
18
|
-
|
|
18
|
+
all_create_argument_lists,
|
|
19
|
+
all_pull_request_argument_lists,
|
|
19
20
|
body_from_arguments,
|
|
20
21
|
create_tool_fields,
|
|
21
22
|
shell_command,
|
|
@@ -34,9 +35,7 @@ from hooks_constants.pull_request_proof_constants import (
|
|
|
34
35
|
AFTER_MERGE_NO_RECORDER_PATTERN,
|
|
35
36
|
AFTER_MERGE_RECORDER_PATTERN,
|
|
36
37
|
AFTER_MERGE_SENTENCE_SEPARATOR_PATTERN,
|
|
37
|
-
ALL_GH_CREATE_WORDS,
|
|
38
38
|
ALL_GH_EDIT_WORDS,
|
|
39
|
-
ALL_PYTHON_PROGRAM_NAMES,
|
|
40
39
|
ALL_VISIBLE_FILE_SUFFIXES,
|
|
41
40
|
CHANGED_FILES_SEPARATOR,
|
|
42
41
|
COMMAND_MARKER,
|
|
@@ -79,9 +78,7 @@ from hooks_constants.pull_request_proof_constants import (
|
|
|
79
78
|
RAW_RENAME_STATUS_LETTER,
|
|
80
79
|
ALL_RAW_TWO_PATH_STATUS_LETTERS,
|
|
81
80
|
PROOF_IMAGE_PATTERN,
|
|
82
|
-
PULL_REQUEST_SCRIPT_CREATE_WORD,
|
|
83
81
|
PULL_REQUEST_SCRIPT_EDIT_WORD,
|
|
84
|
-
PULL_REQUEST_SCRIPT_NAME,
|
|
85
82
|
ALL_TEST_RUNNER_COMMAND_PREFIXES,
|
|
86
83
|
CODE_FENCE_MARKER,
|
|
87
84
|
FENCED_COMMAND_PROMPT,
|
|
@@ -107,7 +104,6 @@ from hooks_constants.pytest_invocation import (
|
|
|
107
104
|
unquoted_token,
|
|
108
105
|
)
|
|
109
106
|
from hooks_constants.shell_command_pipeline import pipeline_segments_for_command
|
|
110
|
-
from hooks_constants.shell_command_segments import token_basename
|
|
111
107
|
from hooks_constants.shell_command_wrappers import segment_program_and_arguments
|
|
112
108
|
|
|
113
109
|
|
|
@@ -191,50 +187,8 @@ def section_under_existing_work(body: str) -> str | None:
|
|
|
191
187
|
return _section_text(body, EXISTING_WORK_HEADING_PATTERN)
|
|
192
188
|
|
|
193
189
|
|
|
194
|
-
def _script_arguments(program: str, all_arguments: list[str]) -> list[str] | None:
|
|
195
|
-
if program == PULL_REQUEST_SCRIPT_NAME:
|
|
196
|
-
return all_arguments
|
|
197
|
-
if program not in ALL_PYTHON_PROGRAM_NAMES or not all_arguments:
|
|
198
|
-
return None
|
|
199
|
-
script_name = token_basename(unquoted_token(all_arguments[0]))
|
|
200
|
-
return all_arguments[1:] if script_name == PULL_REQUEST_SCRIPT_NAME else None
|
|
201
|
-
|
|
202
|
-
|
|
203
|
-
def _script_word_arguments(program: str, all_arguments: list[str], word: str) -> list[str] | None:
|
|
204
|
-
all_script_arguments = _script_arguments(program, all_arguments)
|
|
205
|
-
if all_script_arguments is None or all_script_arguments[:1] != [word]:
|
|
206
|
-
return None
|
|
207
|
-
return all_script_arguments[1:]
|
|
208
|
-
|
|
209
|
-
|
|
210
|
-
def _command_arguments(
|
|
211
|
-
all_segment_tokens: list[str], all_gh_words: list[str], script_word: str
|
|
212
|
-
) -> list[str] | None:
|
|
213
|
-
program, all_arguments = segment_program_and_arguments(all_segment_tokens)
|
|
214
|
-
if program == "gh" and all_arguments[: len(all_gh_words)] == all_gh_words:
|
|
215
|
-
return all_arguments[len(all_gh_words) :]
|
|
216
|
-
return _script_word_arguments(program, all_arguments, script_word)
|
|
217
|
-
|
|
218
|
-
|
|
219
|
-
def _all_argument_lists(command: str, all_gh_words: list[str], script_word: str) -> Iterator[list[str]]:
|
|
220
|
-
all_seen_arguments: set[tuple[str, ...]] = set()
|
|
221
|
-
for each_segment in all_command_segments(command):
|
|
222
|
-
all_arguments = _command_arguments(each_segment, all_gh_words, script_word)
|
|
223
|
-
if all_arguments is None:
|
|
224
|
-
continue
|
|
225
|
-
all_unquoted_arguments = tuple(_unquoted_arguments(all_arguments))
|
|
226
|
-
if all_unquoted_arguments in all_seen_arguments:
|
|
227
|
-
continue
|
|
228
|
-
all_seen_arguments.add(all_unquoted_arguments)
|
|
229
|
-
yield all_arguments
|
|
230
|
-
|
|
231
|
-
|
|
232
|
-
def _all_create_argument_lists(command: str) -> Iterator[list[str]]:
|
|
233
|
-
return _all_argument_lists(command, ALL_GH_CREATE_WORDS, PULL_REQUEST_SCRIPT_CREATE_WORD)
|
|
234
|
-
|
|
235
|
-
|
|
236
190
|
def _all_edit_argument_lists(command: str) -> Iterator[list[str]]:
|
|
237
|
-
return
|
|
191
|
+
return all_pull_request_argument_lists(command, ALL_GH_EDIT_WORDS, PULL_REQUEST_SCRIPT_EDIT_WORD)
|
|
238
192
|
|
|
239
193
|
|
|
240
194
|
def _shell_create_calls(command: str, working_directory: str) -> list[CreateCall]:
|
|
@@ -246,7 +200,7 @@ def _shell_create_calls(command: str, working_directory: str) -> list[CreateCall
|
|
|
246
200
|
_flag_value(all_create_arguments, ALL_BASE_FLAGS, BASE_FLAG_ASSIGNMENT_PREFIX) or None,
|
|
247
201
|
_flag_value(all_create_arguments, ALL_REPOSITORY_FLAGS, REPOSITORY_FLAG_ASSIGNMENT_PREFIX) or None,
|
|
248
202
|
)
|
|
249
|
-
for all_create_arguments in
|
|
203
|
+
for all_create_arguments in all_create_argument_lists(command)
|
|
250
204
|
]
|
|
251
205
|
|
|
252
206
|
|
|
@@ -305,7 +259,7 @@ def is_create_call(all_payload_fields: Mapping[str, object]) -> bool:
|
|
|
305
259
|
"""
|
|
306
260
|
command = shell_command(all_payload_fields)
|
|
307
261
|
return create_tool_fields(all_payload_fields) is not None or (
|
|
308
|
-
command is not None and next(
|
|
262
|
+
command is not None and next(all_create_argument_lists(command), None) is not None
|
|
309
263
|
)
|
|
310
264
|
|
|
311
265
|
|
|
@@ -4,6 +4,7 @@ from __future__ import annotations
|
|
|
4
4
|
|
|
5
5
|
import json
|
|
6
6
|
import sys
|
|
7
|
+
import traceback
|
|
7
8
|
from collections.abc import Iterable, Mapping
|
|
8
9
|
from pathlib import Path
|
|
9
10
|
|
|
@@ -15,6 +16,8 @@ from blocking.codex_apply_patch import payload_patch_target_paths
|
|
|
15
16
|
from blocking.pr_lifecycle_skill_gate import transcript_paths
|
|
16
17
|
from blocking.pull_request_proof import is_create_call
|
|
17
18
|
from blocking.step_note_gate import entry_content_blocks
|
|
19
|
+
from hooks_constants.gate_fault import gate_fault_reason
|
|
20
|
+
from hooks_constants.gate_fault_constants import PROOF_EDIT_GATE_NAME
|
|
18
21
|
from hooks_constants.hook_specific_output_keys import (
|
|
19
22
|
HOOK_EVENT_NAME_KEY,
|
|
20
23
|
HOOK_SPECIFIC_OUTPUT_KEY,
|
|
@@ -158,10 +161,7 @@ def gate_edit_reason(all_payload_fields: Mapping[str, object]) -> str | None:
|
|
|
158
161
|
return None
|
|
159
162
|
|
|
160
163
|
|
|
161
|
-
def
|
|
162
|
-
reason = gate_edit_reason(all_payload_fields)
|
|
163
|
-
if reason is None:
|
|
164
|
-
return None
|
|
164
|
+
def _deny_decision(reason: str) -> dict[str, object]:
|
|
165
165
|
return {
|
|
166
166
|
HOOK_SPECIFIC_OUTPUT_KEY: {
|
|
167
167
|
HOOK_EVENT_NAME_KEY: HOOK_EVENT_NAME,
|
|
@@ -171,8 +171,22 @@ def _gate_edit_decision(all_payload_fields: dict[str, object]) -> dict[str, obje
|
|
|
171
171
|
}
|
|
172
172
|
|
|
173
173
|
|
|
174
|
+
def _gate_edit_decision(all_payload_fields: dict[str, object]) -> dict[str, object] | None:
|
|
175
|
+
try:
|
|
176
|
+
reason = gate_edit_reason(all_payload_fields)
|
|
177
|
+
except Exception as error:
|
|
178
|
+
traceback.print_exc(file=sys.stderr)
|
|
179
|
+
return _deny_decision(gate_fault_reason(PROOF_EDIT_GATE_NAME, error))
|
|
180
|
+
if reason is None:
|
|
181
|
+
return None
|
|
182
|
+
return _deny_decision(reason)
|
|
183
|
+
|
|
184
|
+
|
|
174
185
|
def main() -> int:
|
|
175
|
-
"""Read one Write, Edit, MultiEdit or apply_patch payload and print a deny when it edits the proof gate too soon.
|
|
186
|
+
"""Read one Write, Edit, MultiEdit or apply_patch payload and print a deny when it edits the proof gate too soon.
|
|
187
|
+
|
|
188
|
+
A check that raises also prints a deny, whose reason names the fault.
|
|
189
|
+
"""
|
|
176
190
|
return print_decision_from_stdin(_gate_edit_decision)
|
|
177
191
|
|
|
178
192
|
|
|
@@ -36,6 +36,7 @@ from bash_pre_tool_use_dispatcher import ( # noqa: E402
|
|
|
36
36
|
from hooks_constants.bash_pre_tool_use_dispatcher_constants import ( # noqa: E402
|
|
37
37
|
BASH_TOOL_NAME,
|
|
38
38
|
POWERSHELL_TOOL_NAME,
|
|
39
|
+
BashHostedHookEntry,
|
|
39
40
|
)
|
|
40
41
|
from hooks_constants.hosted_hook_runner import HostedHookRun # noqa: E402
|
|
41
42
|
|
|
@@ -359,3 +360,56 @@ def test_should_stop_the_chain_at_the_first_hosted_hook_that_exits_two(
|
|
|
359
360
|
|
|
360
361
|
assert exit_code == 2
|
|
361
362
|
assert not later_marker.exists()
|
|
363
|
+
|
|
364
|
+
|
|
365
|
+
FAULT_SYSTEM_MESSAGE = "transcript_root_search_gate.py hit a fault and did not check this call"
|
|
366
|
+
|
|
367
|
+
|
|
368
|
+
def test_a_system_message_with_no_decision_is_kept() -> None:
|
|
369
|
+
"""A hosted gate fault prints a systemMessage alone, and aggregation keeps it."""
|
|
370
|
+
fault_run = HostedHookRun(
|
|
371
|
+
captured_stdout=json.dumps({"systemMessage": FAULT_SYSTEM_MESSAGE}) + "\n", did_crash=False
|
|
372
|
+
)
|
|
373
|
+
aggregated = aggregate_bash_hook_results([_run(""), fault_run])
|
|
374
|
+
assert aggregated.decision == ""
|
|
375
|
+
assert aggregated.all_system_messages == [FAULT_SYSTEM_MESSAGE]
|
|
376
|
+
|
|
377
|
+
|
|
378
|
+
def test_dispatch_prints_a_system_message_alone_when_no_hook_decides(
|
|
379
|
+
tmp_path: Path, monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
380
|
+
) -> None:
|
|
381
|
+
"""The dispatcher passes a fault message to the user with no permissionDecision."""
|
|
382
|
+
fault_hook = tmp_path / "fault_hook.py"
|
|
383
|
+
fault_hook.write_text(
|
|
384
|
+
"import json\nprint(json.dumps({'systemMessage': " + repr(FAULT_SYSTEM_MESSAGE) + "}))\n",
|
|
385
|
+
encoding="utf-8",
|
|
386
|
+
)
|
|
387
|
+
monkeypatch.setattr(
|
|
388
|
+
"bash_pre_tool_use_dispatcher.select_applicable_entries",
|
|
389
|
+
lambda tool_name: [
|
|
390
|
+
BashHostedHookEntry(
|
|
391
|
+
script_relative_path=str(fault_hook), applicable_tool_names=frozenset({BASH_TOOL_NAME})
|
|
392
|
+
)
|
|
393
|
+
],
|
|
394
|
+
)
|
|
395
|
+
dispatch(_bash_payload("ls"), BASH_TOOL_NAME)
|
|
396
|
+
all_lines = capsys.readouterr().out.splitlines()
|
|
397
|
+
assert [json.loads(each_line) for each_line in all_lines] == [{"systemMessage": FAULT_SYSTEM_MESSAGE}]
|
|
398
|
+
|
|
399
|
+
|
|
400
|
+
def test_dispatch_prints_nothing_when_every_hook_is_silent(
|
|
401
|
+
tmp_path: Path, monkeypatch: pytest.MonkeyPatch, capsys: pytest.CaptureFixture[str]
|
|
402
|
+
) -> None:
|
|
403
|
+
"""A chain with no decision and no message stays silent."""
|
|
404
|
+
silent_hook = tmp_path / "silent_hook.py"
|
|
405
|
+
silent_hook.write_text("", encoding="utf-8")
|
|
406
|
+
monkeypatch.setattr(
|
|
407
|
+
"bash_pre_tool_use_dispatcher.select_applicable_entries",
|
|
408
|
+
lambda tool_name: [
|
|
409
|
+
BashHostedHookEntry(
|
|
410
|
+
script_relative_path=str(silent_hook), applicable_tool_names=frozenset({BASH_TOOL_NAME})
|
|
411
|
+
)
|
|
412
|
+
],
|
|
413
|
+
)
|
|
414
|
+
dispatch(_bash_payload("ls"), BASH_TOOL_NAME)
|
|
415
|
+
assert capsys.readouterr().out == ""
|
|
@@ -5,15 +5,20 @@ import subprocess
|
|
|
5
5
|
import sys
|
|
6
6
|
from pathlib import Path
|
|
7
7
|
|
|
8
|
+
import pytest
|
|
9
|
+
|
|
8
10
|
HOOKS_DIRECTORY = Path(__file__).resolve().parent.parent
|
|
9
11
|
if str(HOOKS_DIRECTORY) not in sys.path:
|
|
10
12
|
sys.path.insert(0, str(HOOKS_DIRECTORY))
|
|
11
13
|
|
|
12
14
|
from blocking import build_eval_before_eval_write_gate as gate
|
|
15
|
+
from blocking.gate_run_reads import ReadSlice, read_deadlines_started, started_read_deadlines
|
|
13
16
|
from hooks_constants.build_eval_before_eval_write_gate_constants import (
|
|
17
|
+
EVAL_WRITE_SCAN_RAN_OUT_SENTENCE,
|
|
14
18
|
MISSING_BUILD_EVAL_BEFORE_EVAL_WRITE_REASON,
|
|
15
19
|
UNREADABLE_TRANSCRIPT_BEFORE_EVAL_WRITE_REASON,
|
|
16
20
|
)
|
|
21
|
+
from hooks_constants.gate_run_reads_constants import EVAL_WRITE_SCAN_SECONDS, STARTUP_AND_OUTPUT_MARGIN_SECONDS
|
|
17
22
|
|
|
18
23
|
GATE_SCRIPT = Path(gate.__file__)
|
|
19
24
|
|
|
@@ -160,3 +165,62 @@ def test_unparseable_apply_patch_naming_an_eval_file_should_be_denied(tmp_path:
|
|
|
160
165
|
"transcript_path": str(_transcript(tmp_path, build_eval=False)),
|
|
161
166
|
}
|
|
162
167
|
assert _reason(gate.decision_for(payload)) == MISSING_BUILD_EVAL_BEFORE_EVAL_WRITE_REASON
|
|
168
|
+
|
|
169
|
+
|
|
170
|
+
EVAL_WRITE_GATE_COMMAND_SUFFIX = "/hooks/blocking/build_eval_before_eval_write_gate.py"
|
|
171
|
+
CLOCK_START = 1000.0
|
|
172
|
+
|
|
173
|
+
|
|
174
|
+
class SteppedClock:
|
|
175
|
+
def __init__(self) -> None:
|
|
176
|
+
self.now = CLOCK_START
|
|
177
|
+
|
|
178
|
+
def __call__(self) -> float:
|
|
179
|
+
return self.now
|
|
180
|
+
|
|
181
|
+
|
|
182
|
+
def _registered_gate_timeout_seconds() -> int:
|
|
183
|
+
all_registration_fields = json.loads((HOOKS_DIRECTORY / "hooks.json").read_text(encoding="utf-8"))
|
|
184
|
+
all_timeouts = [
|
|
185
|
+
each_hook["timeout"]
|
|
186
|
+
for each_group in all_registration_fields["hooks"]["PreToolUse"]
|
|
187
|
+
for each_hook in each_group["hooks"]
|
|
188
|
+
if each_hook.get("command", "").endswith(EVAL_WRITE_GATE_COMMAND_SUFFIX)
|
|
189
|
+
]
|
|
190
|
+
assert len(all_timeouts) == 1
|
|
191
|
+
return all_timeouts[0]
|
|
192
|
+
|
|
193
|
+
|
|
194
|
+
def test_registered_timeout_should_outlast_the_eval_write_scan_deadline() -> None:
|
|
195
|
+
assert ReadSlice.EVAL_WRITE.value == EVAL_WRITE_SCAN_SECONDS
|
|
196
|
+
assert _registered_gate_timeout_seconds() >= EVAL_WRITE_SCAN_SECONDS + STARTUP_AND_OUTPUT_MARGIN_SECONDS
|
|
197
|
+
|
|
198
|
+
|
|
199
|
+
def test_scan_the_eval_write_deadline_stopped_should_deny_and_say_to_call_build_eval_again(tmp_path: Path) -> None:
|
|
200
|
+
clock = SteppedClock()
|
|
201
|
+
payload = _payload("/repo/evals/cases.json", _transcript(tmp_path, build_eval=True))
|
|
202
|
+
with read_deadlines_started(clock):
|
|
203
|
+
clock.now = CLOCK_START + EVAL_WRITE_SCAN_SECONDS
|
|
204
|
+
decision = gate.decision_for(payload)
|
|
205
|
+
assert _reason(decision) == MISSING_BUILD_EVAL_BEFORE_EVAL_WRITE_REASON + EVAL_WRITE_SCAN_RAN_OUT_SENTENCE
|
|
206
|
+
|
|
207
|
+
|
|
208
|
+
def test_scan_with_time_left_should_deny_without_the_ran_out_sentence(tmp_path: Path) -> None:
|
|
209
|
+
clock = SteppedClock()
|
|
210
|
+
payload = _payload("/repo/evals/cases.json", _transcript(tmp_path, build_eval=False))
|
|
211
|
+
with read_deadlines_started(clock):
|
|
212
|
+
clock.now = CLOCK_START + EVAL_WRITE_SCAN_SECONDS - 1
|
|
213
|
+
assert _reason(gate.decision_for(payload)) == MISSING_BUILD_EVAL_BEFORE_EVAL_WRITE_REASON
|
|
214
|
+
assert gate.decision_for(_payload("/repo/evals/cases.json", _transcript(tmp_path, build_eval=True))) is None
|
|
215
|
+
|
|
216
|
+
|
|
217
|
+
def test_main_should_start_the_read_deadlines_before_it_reads_the_payload(monkeypatch: pytest.MonkeyPatch) -> None:
|
|
218
|
+
all_seen_deadlines: list[object] = []
|
|
219
|
+
|
|
220
|
+
def recorded_print(decide: object) -> int:
|
|
221
|
+
all_seen_deadlines.append(started_read_deadlines.get())
|
|
222
|
+
return 0
|
|
223
|
+
|
|
224
|
+
monkeypatch.setattr(gate, "print_decision_from_stdin", recorded_print)
|
|
225
|
+
assert gate.main() == 0
|
|
226
|
+
assert all_seen_deadlines[0] is not None
|