@simplysm/sd-claude 13.0.75 → 13.0.77

package/claude/skills/sd-skill/SKILL.md

@@ -176,12 +176,23 @@ Edit skill without testing? Same violation.
 - Don't "adapt" while running tests
 - Delete means delete
 
+**Only exemption — pure mechanical edits:** Typo fixes, tool/variable renames where the behavioral guidance is identical (e.g., `TodoWrite` → `TaskCreate`). If you're changing what the skill *teaches*, it's not mechanical — test it.
+
 **REQUIRED BACKGROUND:** The sd-tdd skill explains why this matters. Same principles apply to documentation.
 
 ## Testing All Skill Types
 
 Different skill types need different test approaches:
 
+```mermaid
+flowchart TD
+    A{"What type of skill?"}
+    A -->|"Discipline (rules/requirements)"| B["Pressure test<br>(compliance under stress)"]
+    A -->|"Technique (how-to guides)"| C["Application test<br>(correct technique usage)"]
+    A -->|"Pattern (mental models)"| D["Recognition test<br>(when/how to apply)"]
+    A -->|"Reference (docs/APIs)"| E["Retrieval test<br>(find & use reference)"]
+```
+
 ### Discipline-Enforcing Skills (rules/requirements)
 
 **Examples:** TDD, verification-before-completion, designing-before-coding
@@ -253,6 +264,8 @@ Example: Testing a "condition-based-waiting" skill
 | "Academic review is enough"            | Reading ≠ using. Test application scenarios.                                                                    |
 | "No time to test"                      | Deploying untested skill wastes more time fixing it later.                                                      |
 | "I already know the baseline failures" | You know what YOU think the failures are. Run a subagent to see what ACTUALLY happens. Knowledge ≠ observation. |
+| "This is process theater" | If the process catches even one issue you missed, it paid for itself. "Theater" is what you call process before it saves you. |
+| "It applies the wrong test methodology" | Different skill types need different tests (pressure vs retrieval), but ALL types need testing. No type is exempt. |
 
 **All of these mean: Test before deploying. No exceptions.**
 
@@ -264,6 +277,10 @@ Skills that enforce discipline need to resist rationalization. **See writing-gui
 
 Follow the TDD cycle:
 
+### Subagent Rules
+
+**NEVER use `isolation: "worktree"` when launching subagents.** Worktrees break lint/build tooling. Always run subagents in the default (non-isolated) mode.
+
 ### RED: Write Failing Test (Baseline)
 
 Run pressure scenario with subagent WITHOUT the skill. Document exact behavior:
@@ -307,9 +324,9 @@ example-js.js, example-py.py, example-go.go
 
 ### ❌ Code in Flowcharts
 
-```dot
-step1 [label="import fs"];
-step2 [label="read file"];
+```mermaid
+flowchart TD
+    A["import fs"] --> B["read file"]
 ```
 
 **Why bad:** Can't copy-paste, hard to read