claude-dev-env 8.98.4 → 8.98.7

package/hooks/test_transcript_skill_scan.py

@@ -9,10 +9,12 @@ if str(HOOKS_DIRECTORY) not in sys.path:
     sys.path.insert(0, str(HOOKS_DIRECTORY))
 
 from transcript_skill_scan import (
+    TranscriptToolCall,
     invokes_skill_with_argument,
     is_skill_loaded_after_last_compaction,
     is_skill_loaded_since_last_prompt,
     skill_invocation_status,
+    tool_calls_since_last_prompt,
 )
 
 
@@ -158,3 +160,31 @@ def test_slash_command_prompt_loads_the_skill_for_its_own_turn() -> None:
     assert is_skill_loaded_since_last_prompt([_prompt_entry("First."), _prompt_entry(marker)], ("verify",), (marker,))
     compaction = json.dumps({"subtype": "compact_boundary"})
     assert not is_skill_loaded_since_last_prompt([_prompt_entry(marker), compaction], ("verify",), (marker,))
+
+
+def should_list_each_tool_call_since_the_last_prompt_with_its_outcome() -> None:
+    def _call(tool_use_id: str, name: str) -> str:
+        return json.dumps(
+            {"type": "assistant", "message": {"content": [{"type": "tool_use", "id": tool_use_id, "name": name, "input": {}}]}}
+        )
+
+    def _result(tool_use_id: str, is_error: bool) -> str:
+        tool_result = {"type": "tool_result", "tool_use_id": tool_use_id, "is_error": is_error, "content": "x"}
+        return json.dumps({"type": "user", "message": {"content": [tool_result]}})
+
+    lines = [
+        json.dumps({"type": "user", "message": {"content": "First."}}),
+        _call("old", "Read"),
+        _result("old", False),
+        json.dumps({"type": "user", "message": {"content": "Second."}}),
+        _call("ok", "Bash"),
+        _result("ok", False),
+        _call("denied", "Edit"),
+        _result("denied", True),
+        _call("pending", "Write"),
+    ]
+    assert tool_calls_since_last_prompt(lines) == [
+        TranscriptToolCall("Bash", {}, True),
+        TranscriptToolCall("Edit", {}, False),
+        TranscriptToolCall("Write", {}, False),
+    ]

package/hooks/transcript_skill_scan.py

@@ -1,23 +1,37 @@
-"""Find a skill invocation after the last transcript compaction, or one with a given first argument anywhere."""
+"""Find a skill invocation after the last transcript compaction, or one with a given first argument anywhere.
+
+It also lists the tool calls made since the last user prompt, each with whether its result came back without an error.
+"""
 
 from __future__ import annotations
 
 import json
 import re
 from collections.abc import Collection, Iterable, Iterator
+from typing import NamedTuple
 
 from hooks_constants.transcript_skill_scan_constants import (
     ASSISTANT_ENTRY_TYPE,
     COMMAND_ARGUMENTS_PATTERN,
     COMPACT_BOUNDARY_SUBTYPE,
+    IS_ERROR_FIELD,
     META_ENTRY_FIELD,
     SKILL_TOOL_NAME,
     TOOL_RESULT_BLOCK_TYPE,
     TOOL_USE_BLOCK_TYPE,
+    TOOL_USE_ID_FIELD,
     USER_ENTRY_TYPE,
 )
 
 
+class TranscriptToolCall(NamedTuple):
+    """One tool call in a transcript: its name, its input, and whether its result came back without an error."""
+
+    name: str
+    tool_input: dict[str, object]
+    succeeded: bool
+
+
 def _user_command_text(all_entry_fields: dict[str, object]) -> str | None:
     message = all_entry_fields.get("message")
     if all_entry_fields.get("type") != USER_ENTRY_TYPE or not isinstance(message, dict):
@@ -225,3 +239,60 @@ def invokes_skill_with_argument(
         if all_entry_fields and _invokes_skill_with_argument(all_entry_fields, skill_names, argument_word, command_marker):
             return True
     return False
+
+
+def _content_blocks(all_entry_fields: dict[str, object], entry_type: str) -> list[dict[str, object]]:
+    message = all_entry_fields.get("message")
+    if all_entry_fields.get("type") != entry_type or not isinstance(message, dict):
+        return []
+    content = message.get("content")
+    return [each_block for each_block in (content if isinstance(content, list) else []) if isinstance(each_block, dict)]
+
+
+def _tool_use_calls(all_entry_fields: dict[str, object]) -> list[tuple[object, str, dict[str, object]]]:
+    all_tool_use_calls: list[tuple[object, str, dict[str, object]]] = []
+    for each_block in _content_blocks(all_entry_fields, ASSISTANT_ENTRY_TYPE):
+        tool_name = each_block.get("name")
+        tool_input = each_block.get("input")
+        if each_block.get("type") == TOOL_USE_BLOCK_TYPE and isinstance(tool_name, str):
+            all_tool_use_calls.append((each_block.get("id"), tool_name, tool_input if isinstance(tool_input, dict) else {}))
+    return all_tool_use_calls
+
+
+def _succeeded_result_ids(all_entry_fields: dict[str, object]) -> set[object]:
+    return {
+        each_block.get(TOOL_USE_ID_FIELD)
+        for each_block in _content_blocks(all_entry_fields, USER_ENTRY_TYPE)
+        if each_block.get("type") == TOOL_RESULT_BLOCK_TYPE and each_block.get(IS_ERROR_FIELD) is not True
+    }
+
+
+def tool_calls_since_last_prompt(all_transcript_lines: Iterable[str]) -> list[TranscriptToolCall]:
+    """Return the tool calls after the last user prompt, in order, each marked by whether it succeeded.
+
+    ::
+
+        prompt, reply call, its result without an error  -> [reply succeeded]
+        prompt, reply call, its result with is_error     -> [reply failed]
+        prompt, resolve call with no result yet          -> [resolve failed]
+        reply call and result, then a new prompt         -> []
+
+    A hook that denies a call leaves an errored result, so a denied call does
+    not count as a success. A tool result leaves the turn open.
+
+    Args:
+        all_transcript_lines: JSON transcript entries, one per line.
+    """
+    all_calls: list[tuple[object, str, dict[str, object]]] = []
+    all_succeeded_ids: set[object] = set()
+    all_entries = (each_entry for each_entry in map(_parsed_entry, all_transcript_lines) if each_entry is not None)
+    for each_entry_fields in all_entries:
+        if _is_user_prompt(each_entry_fields):
+            all_calls, all_succeeded_ids = [], set()
+            continue
+        all_calls.extend(_tool_use_calls(each_entry_fields))
+        all_succeeded_ids.update(_succeeded_result_ids(each_entry_fields))
+    return [
+        TranscriptToolCall(each_name, each_input, each_id is not None and each_id in all_succeeded_ids)
+        for each_id, each_name, each_input in all_calls
+    ]

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "claude-dev-env",
-    "version": "8.98.4",
+    "version": "8.98.7",
     "description": "Claude Code development standards — rules, hooks, agents, commands, and skills",
     "type": "module",
     "bin": {

package/scripts/durable_post_lint.py

@@ -18,6 +18,7 @@ from dev_env_scripts_constants.private_term_constants import (
     PRIVATE_TERM_MESSAGE_TEMPLATE,
 )
 from durable_post_lint_config.config.constants import (
+    ACTION_GITHUB_MCP_COMMENT,
     ACTION_PR_CREATE,
     ACTION_PR_EDIT,
     ALL_BARE_VOLATILE_PATH_MARKERS,
@@ -33,6 +34,7 @@ from durable_post_lint_config.config.constants import (
     BODY_MUST_NOT_BE_EMPTY_MESSAGE,
     BODY_REQUIRED_MESSAGE,
     CLEAN_EXIT_CODE,
+    COMMENT_LENGTH_FINDING_CODE,
     CONTENT_FINDING_EXIT_CODE,
     CONVENTIONAL_TITLE_PATTERN,
     EDIT_INPUT_REQUIRED_MESSAGE,
@@ -45,6 +47,8 @@ from durable_post_lint_config.config.constants import (
     MISSING_HEADING_MESSAGE_TEMPLATE,
     PATH_ANCHOR_CHARACTER,
     PATH_SEGMENT_START_CHARACTERS,
+    PROOF_GATE_CREATE_TOOL_NAME,
+    PULL_REQUEST_PROOF_FINDING_CODE,
     RELEASE_BRANCH_PREFIX,
     REPOSITORY_OWNER_SEPARATOR,
     REWRITTEN_RELEASE_BODY_CODE,
@@ -56,6 +60,17 @@ from durable_post_lint_config.config.constants import (
 )
 from private_terms import private_term_line_numbers
 
+_hooks_directory = str(Path(__file__).resolve().parents[1] / "hooks")
+if _hooks_directory not in sys.path:
+    sys.path.insert(0, _hooks_directory)
+
+from blocking.pull_request_proof import (
+    missing_eval_section_reason,
+    missing_existing_work_reason,
+    missing_proof_reason,
+)
+from blocking.reply_length_gate import comment_length_violation
+
 
 class DurablePostUsageError(ValueError):
     """Report an invalid GitHub post request shape."""
@@ -145,6 +160,32 @@ def _description_heading_findings(body_text: str) -> list[DurablePostFinding]:
     ]
 
 
+def _pull_request_proof_findings(
+    title: str | None, body_text: str
+) -> list[DurablePostFinding]:
+    create_payload = {
+        "tool_name": PROOF_GATE_CREATE_TOOL_NAME,
+        "tool_input": {"title": title, "body": body_text},
+    }
+    all_reasons = (
+        missing_proof_reason(create_payload),
+        missing_existing_work_reason(create_payload),
+        missing_eval_section_reason(create_payload),
+    )
+    return [
+        DurablePostFinding(code=PULL_REQUEST_PROOF_FINDING_CODE, message=each_reason)
+        for each_reason in all_reasons
+        if each_reason is not None
+    ]
+
+
+def _comment_length_findings(body_text: str) -> list[DurablePostFinding]:
+    violation = comment_length_violation(body_text)
+    if violation is None:
+        return []
+    return [DurablePostFinding(code=COMMENT_LENGTH_FINDING_CODE, message=violation)]
+
+
 def _contrast_framing_findings(text: str) -> list[DurablePostFinding]:
     all_lines = text.splitlines()
     return [
@@ -210,6 +251,7 @@ def _body_findings(
     action: str,
     body_text: str | None,
     head_branch: str | None = None,
+    title: str | None = None,
 ) -> list[DurablePostFinding]:
     if body_text is None:
         return []
@@ -226,6 +268,10 @@ def _body_findings(
     else:
         if action in ALL_PR_DESCRIPTION_ACTIONS:
             all_findings.extend(_description_heading_findings(body_text))
+        if action == ACTION_PR_CREATE:
+            all_findings.extend(_pull_request_proof_findings(title, body_text))
+        if action == ACTION_GITHUB_MCP_COMMENT:
+            all_findings.extend(_comment_length_findings(body_text))
         all_findings.extend(_contrast_framing_findings(body_text))
     if find_volatile_path_marker(body_text) is not None:
         all_findings.append(
@@ -263,7 +309,7 @@ def lint_durable_post(
     return (
         *_title_findings(title),
         *_private_term_findings(title, repository),
-        *_body_findings(action, body_text, head_branch),
+        *_body_findings(action, body_text, head_branch, title),
         *_private_term_findings(body_text, repository),
     )
 

package/scripts/durable_post_lint_config/config/constants.py

@@ -2,6 +2,7 @@
 
 import re
 
+ACTION_GITHUB_MCP_COMMENT = "github-mcp-comment"
 ACTION_GITHUB_MCP_POST = "github-mcp-post"
 ACTION_ISSUE_COMMENT = "issue-comment"
 ACTION_ISSUE_CREATE = "issue-create"
@@ -13,6 +14,7 @@ ACTION_PR_REVIEW = "pr-review"
 
 ALL_POST_ACTIONS: frozenset[str] = frozenset(
     {
+        ACTION_GITHUB_MCP_COMMENT,
         ACTION_GITHUB_MCP_POST,
         ACTION_ISSUE_COMMENT,
         ACTION_ISSUE_CREATE,
@@ -25,6 +27,7 @@ ALL_POST_ACTIONS: frozenset[str] = frozenset(
 )
 ALL_BODY_REQUIRED_ACTIONS: frozenset[str] = frozenset(
     {
+        ACTION_GITHUB_MCP_COMMENT,
         ACTION_GITHUB_MCP_POST,
         ACTION_ISSUE_COMMENT,
         ACTION_ISSUE_CREATE,
@@ -102,3 +105,6 @@ REWRITTEN_RELEASE_BODY_MESSAGE = (
     "keep the generated text whole"
 )
 REPOSITORY_OWNER_SEPARATOR = "/"
+PROOF_GATE_CREATE_TOOL_NAME = "mcp__github__create_pull_request"
+PULL_REQUEST_PROOF_FINDING_CODE = "pull-request-proof-section"
+COMMENT_LENGTH_FINDING_CODE = "github-comment-length"

package/scripts/durable_post_lint_config/config/test_constants.py

@@ -12,6 +12,7 @@ from durable_post_lint_config.config.constants import (
 
 def test_action_inventory_covers_pull_request_issue_and_tool_posts() -> None:
     assert ALL_POST_ACTIONS == {
+        "github-mcp-comment",
         "github-mcp-post",
         "issue-comment",
         "issue-create",

package/scripts/policy_lint/adapter_configuration.py

@@ -144,6 +144,7 @@ def _names_exempt_registration_path(registered_string: str) -> bool:
         hooks/blocking/open_tasks_stop_gate.py           -> exempt
         hooks/blocking/owned_pr_merge_stop_gate.py       -> exempt
         hooks/blocking/session_title_turn_gate.py        -> exempt
+        hooks/blocking/resolve_thread_reply_gate.py      -> exempt
         hooks/blocking/some_new_blocker.py               -> flagged
 
     The Bash PreToolUse dispatcher sits under ``blocking/`` for layout reasons
@@ -202,6 +203,10 @@ def _names_exempt_registration_path(registered_string: str) -> bool:
     closing reply names no ``Blocked:`` line for it. It decides when a turn may
     end and no code or safety policy.
 
+    The resolve reply gate denies a review thread resolve until a reply on the
+    thread has posted. It decides the order of two review calls and no code or
+    safety policy.
+
     Args:
         registered_string: One command, path, script, or entrypoint string.
     """

package/scripts/policy_lint/config/constants.py

@@ -83,6 +83,7 @@ ALL_ACTION_BOUNDARY_EXEMPT_REGISTRATION_PATHS = (
     "blocking/open_tasks_stop_gate.py",
     "blocking/owned_pr_merge_stop_gate.py",
     "blocking/session_title_turn_gate.py",
+    "blocking/resolve_thread_reply_gate.py",
 )
 ALL_TEST_DIRECTORY_NAMES = frozenset({"tests"})
 ALL_TEST_FILE_PREFIXES = ("test_",)

package/scripts/test_durable_post_lint.py

@@ -11,6 +11,11 @@ import pytest
 
 import durable_post_lint
 from dev_env_scripts_constants.private_term_constants import PrivateTermDigest
+from hooks_constants.pull_request_proof_constants import (
+    MISSING_EVAL_SECTION_REASON,
+    MISSING_EXISTING_WORK_REASON,
+    TEST_ONLY_PROOF_REASON,
+)
 
 SCRIPT_PATH = Path(__file__).resolve().parent / "durable_post_lint.py"
 VALID_PR_BODY = """## Why
@@ -32,7 +37,25 @@ Blast radius text.
 ## Verification
 
 Verification text.
+
+## Existing work
+
+`scripts/durable_post_lint.py:1` holds the post checks.
+
+## Proof in practice
+
+Ran `python scripts/durable_post_lint.py --action pr-create --body-file body.md` on the new body.
+
+## Eval
+
+`python scripts/durable_post_lint.py --action pr-create --body-file eval_body.md`
 """
+OVER_LONG_REPLY_BODY = (
+    "Fixed in `abc1234`.\n\n"
+    "The parser now rejects an empty header.\n\n"
+    "The test covers both cases.\n\n"
+    "---\nPosted by the review agent."
+)
 MINIMAL_PR_BODY = """## Why
 
 Why text.
@@ -54,6 +77,7 @@ Verification text.
         ("issue-edit", None, "Updated issue body."),
         ("issue-comment", None, "Issue comment."),
         ("github-mcp-post", None, "MCP body."),
+        ("github-mcp-comment", None, "Reply body."),
     ],
 )
 def test_all_actions_accept_valid_inputs(
@@ -112,6 +136,7 @@ def test_pr_edit_requires_one_edit_field() -> None:
         "issue-edit",
         "issue-comment",
         "github-mcp-post",
+        "github-mcp-comment",
     ],
 )
 def test_body_actions_require_body(action: str) -> None:
@@ -176,10 +201,9 @@ def test_pr_body_reports_each_missing_heading(missing_heading: str) -> None:
     ]
 
 
-@pytest.mark.parametrize("action", ["pr-create", "pr-edit"])
-def should_accept_a_body_carrying_only_the_two_required_headings(action: str) -> None:
+def should_accept_an_edit_body_carrying_only_the_two_required_headings() -> None:
     all_findings = durable_post_lint.lint_durable_post(
-        action=action,
+        action="pr-edit",
         title="fix(cli): tighten input handling",
         body_text=MINIMAL_PR_BODY,
     )
@@ -201,6 +225,155 @@ def should_still_accept_a_body_published_under_the_superseded_heading_set() -> N
     assert all_findings == ()
 
 
+def should_report_a_create_body_with_no_existing_work_section() -> None:
+    body_text = VALID_PR_BODY.replace(
+        "## Existing work\n\n`scripts/durable_post_lint.py:1` holds the post checks.\n\n", ""
+    )
+    all_findings = durable_post_lint.lint_durable_post(
+        action="pr-create",
+        title="feat: add durable post linting",
+        body_text=body_text,
+    )
+    assert [each_finding.message for each_finding in all_findings] == [
+        MISSING_EXISTING_WORK_REASON
+    ]
+
+
+def should_report_a_create_body_whose_proof_runs_only_tests() -> None:
+    body_text = VALID_PR_BODY.replace(
+        "Ran `python scripts/durable_post_lint.py --action pr-create --body-file body.md` on the new body.",
+        "Ran `python -m pytest scripts`.",
+    )
+    all_findings = durable_post_lint.lint_durable_post(
+        action="pr-create",
+        title="fix: tighten durable post linting",
+        body_text=body_text,
+    )
+    assert [each_finding.message for each_finding in all_findings] == [
+        TEST_ONLY_PROOF_REASON
+    ]
+
+
+def should_report_a_feature_create_body_with_no_eval_section() -> None:
+    body_text = VALID_PR_BODY.split("## Eval")[0]
+    all_findings = durable_post_lint.lint_durable_post(
+        action="pr-create",
+        title="feat: add durable post linting",
+        body_text=body_text,
+    )
+    assert [each_finding.message for each_finding in all_findings] == [
+        MISSING_EVAL_SECTION_REASON
+    ]
+
+
+def should_leave_an_edit_body_with_no_existing_work_section_clean() -> None:
+    all_findings = durable_post_lint.lint_durable_post(
+        action="pr-edit",
+        title=None,
+        body_text=MINIMAL_PR_BODY,
+    )
+    assert all_findings == ()
+
+
+def should_exit_with_findings_for_a_create_body_with_no_existing_work_section(
+    tmp_path: Path,
+) -> None:
+    body_file = tmp_path / "body.md"
+    body_file.write_text(
+        VALID_PR_BODY.replace("## Existing work", "## Prior art"), encoding="utf-8"
+    )
+    completed_process = subprocess.run(
+        [
+            sys.executable,
+            str(SCRIPT_PATH),
+            "--action",
+            "pr-create",
+            "--title",
+            "feat: add durable post linting",
+            "--body-file",
+            str(body_file),
+        ],
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+    assert completed_process.returncode == 1
+    assert completed_process.stderr == f"{MISSING_EXISTING_WORK_REASON}\n"
+
+
+def should_report_a_review_reply_over_three_sentences() -> None:
+    all_findings = durable_post_lint.lint_durable_post(
+        action="github-mcp-comment",
+        title=None,
+        body_text=OVER_LONG_REPLY_BODY,
+    )
+    assert [each_finding.code for each_finding in all_findings] == [
+        "github-comment-length"
+    ]
+
+
+def should_report_a_review_reply_sentence_over_fifteen_words() -> None:
+    all_findings = durable_post_lint.lint_durable_post(
+        action="github-mcp-comment",
+        title=None,
+        body_text=(
+            "This change moves the parser into its own module and keeps every"
+            " caller on the same import path."
+        ),
+    )
+    assert [each_finding.code for each_finding in all_findings] == [
+        "github-comment-length"
+    ]
+
+
+def should_accept_a_short_review_reply() -> None:
+    all_findings = durable_post_lint.lint_durable_post(
+        action="github-mcp-comment",
+        title=None,
+        body_text="Fixed in `abc1234`.\n\nThe parser now rejects an empty header.",
+    )
+    assert all_findings == ()
+
+
+def should_exit_with_findings_for_an_over_long_review_reply(tmp_path: Path) -> None:
+    body_file = tmp_path / "reply.md"
+    body_file.write_text(OVER_LONG_REPLY_BODY, encoding="utf-8")
+    completed_process = subprocess.run(
+        [
+            sys.executable,
+            str(SCRIPT_PATH),
+            "--action",
+            "github-mcp-comment",
+            "--body-file",
+            str(body_file),
+        ],
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+    assert completed_process.returncode == 1
+    assert "at most three short sentences" in completed_process.stderr
+
+
+def should_exit_clean_for_a_short_review_reply(tmp_path: Path) -> None:
+    body_file = tmp_path / "reply.md"
+    body_file.write_text("Fixed in `abc1234`.\n", encoding="utf-8")
+    completed_process = subprocess.run(
+        [
+            sys.executable,
+            str(SCRIPT_PATH),
+            "--action",
+            "github-mcp-comment",
+            "--body-file",
+            str(body_file),
+        ],
+        check=False,
+        capture_output=True,
+        text=True,
+    )
+    assert completed_process.returncode == 0
+
+
 def test_non_pr_body_needs_no_pr_headings() -> None:
     all_findings = durable_post_lint.lint_durable_post(
         action="issue-comment",

package/scripts/tests/test_adapter_configuration.py

@@ -58,6 +58,14 @@ def test_pull_request_lifecycle_gate_registration_is_clean(tmp_path: Path) -> No
     assert adapters.hook_configuration_diagnostics(current_document, tmp_path) == ()
 
 
+def test_new_resolve_thread_reply_gate_registration_is_clean(tmp_path: Path) -> None:
+    current_document = _hook_document(
+        ["hooks/blocking/resolve_thread_reply_gate.py"],
+        '{"hooks": {}}',
+    )
+    assert adapters.hook_configuration_diagnostics(current_document, tmp_path) == ()
+
+
 def test_new_merge_offer_gate_registration_is_clean(tmp_path: Path) -> None:
     current_document = _hook_document(
         ["hooks/blocking/merge_over_review_offer_gate.py"],