fix(integration): address reviewer portability and guidance

docs/TROUBLESHOOTING.md

@@ -62,7 +62,7 @@ exit 2
 **What helps:**
 
 - Reduce tool-definition pressure with `ENABLE_TOOL_SEARCH=auto:5` if your setup supports it.
-- Lower the supported effort level (for example, `/effort low`) for routine work on adaptive-reasoning models; `MAX_THINKING_TOKENS` only caps fixed-budget thinking.
+- Lower the supported effort level (for example, `/effort low`) for routine work on adaptive-reasoning models. `MAX_THINKING_TOKENS=0` disables thinking; nonzero values cap fixed-budget thinking only when adaptive thinking is disabled.
 - Route subagent work to a cheaper model such as `CLAUDE_CODE_SUBAGENT_MODEL=haiku` if your setup exposes that knob.
 - Disable unused MCP servers per project.
 - Compact manually at natural breakpoints instead of waiting for auto-compaction.

docs/hook-bug-workarounds.md

@@ -61,7 +61,7 @@ What helps:
 What helps:
 
 - Reduce tool-definition pressure with `ENABLE_TOOL_SEARCH=auto:5` if your setup supports it.
-- Lower the supported effort level (for example, `/effort low`) for routine work on adaptive-reasoning models; `MAX_THINKING_TOKENS` only caps fixed-budget thinking.
+- Lower the supported effort level (for example, `/effort low`) for routine work on adaptive-reasoning models. `MAX_THINKING_TOKENS=0` disables thinking; nonzero values cap fixed-budget thinking only when adaptive thinking is disabled.
 - Route subagent work to a cheaper model such as `CLAUDE_CODE_SUBAGENT_MODEL=haiku` if your setup exposes that knob.
 - Disable unused MCP servers per project.
 - Compact manually at natural breakpoints instead of waiting for auto-compaction.

skills/continuous-learning-v2/scripts/test_parse_instinct.py

@@ -281,6 +281,7 @@ def test_parse_confidence_preserves_valid_range(confidence):
     assert parse_instinct_file(content)[0]["confidence"] == confidence
 
 
+@pytest.mark.unit
 @pytest.mark.parametrize("value", [
     "line one\nline two", "tab\tvalue", "carriage\rreturn",
     'when writing "quoted" text', r"C:\Users\name", "caf\u00e9",
@@ -294,11 +295,13 @@ def test_quoted_instinct_strings_roundtrip(value: str) -> None:
     assert instincts[0]["trigger"] == value
 
 
+@pytest.mark.unit
 def test_parse_quoted_instinct_decodes_newline_escape() -> None:
     content = '---\nid: newline\ntrigger: "line one\\nline two"\n---\nBody'
     assert parse_instinct_file(content)[0]["trigger"] == "line one\nline two"
 
 
+@pytest.mark.unit
 @pytest.mark.parametrize("value", ["line\u0085value", "line\u0090value"])
 def test_unicode_controls_are_escaped_for_yaml(value: str) -> None:
     quoted = _mod._yaml_quote(value)
@@ -306,6 +309,7 @@ def test_unicode_controls_are_escaped_for_yaml(value: str) -> None:
     assert json.loads(quoted) == value
 
 
+@pytest.mark.unit
 def test_parse_quoted_instinct_preserves_legacy_escape() -> None:
     content = '---\nid: legacy\ntrigger: "C:\\projects"\n---\nBody'
     assert parse_instinct_file(content)[0]["trigger"] == r"C:\projects"

tests/lib/resolve-ecc-root.test.js

@@ -548,7 +548,7 @@ module.exports = { resolveEccRoot() { assert.strictEqual(process.env.HOME, ${JSO
         });
         // On WSL, 'C:\\Users\\test\\.claude' is converted to '/mnt/c/Users/test/.claude'.
         // No install exists at the converted path, so resolveEccRoot falls back to it.
-        assert.strictEqual(result, '/mnt/c/Users/test/.claude');
+        assert.strictEqual(result, path.resolve('/mnt/c/Users/test/.claude'));
       } finally {
         fs.rmSync(homeDir, { recursive: true, force: true });
       }