@fr0mpy/prompt-system 2.0.0

package/templates/rules/subagent-format.md

@@ -0,0 +1,94 @@
+<!-- SUMMARY: Follow standard format when creating or modifying subagents -->
+<!-- TRIGGER: subagent -->
+# RULE: Subagent Format
+
+⚠️ **ACTIVE RULE** - Apply when creating or modifying subagent files.
+
+## 🔐 Enforcement
+
+**When creating/modifying subagents, you MUST:**
+1. Use YAML frontmatter with required fields
+2. Include structured output format for context return
+3. Keep prompts concise and actionable
+4. Place in `.claude/agents/` directory
+
+---
+
+## Template
+
+```markdown
+---
+name: agent-name
+description: Brief description. When to use this agent proactively.
+tools: Grep, Glob, Read, WebSearch, Bash
+model: haiku
+---
+
+You are a [role] for [project type].
+
+## Your Task
+
+1. Step one
+2. Step two
+3. Step three
+
+## Output Format
+
+Return findings in this EXACT structure for context handoff:
+
+### RESULT: [found | not-found | partial]
+
+### FILES FOUND:
+- `path/to/file.ts:line` - [what's there]
+- `path/to/other.ts:line` - [what's there]
+
+### PATTERNS DETECTED:
+- [existing pattern 1]
+- [existing pattern 2]
+
+### GAPS:
+- [what's missing]
+
+### RECOMMENDATION:
+[Specific action to take based on findings]
+
+## Rules
+
+- Rule one
+- Rule two
+```
+
+---
+
+## Required Fields
+
+| Field | Purpose |
+|-------|---------|
+| `name` | Unique identifier (kebab-case) |
+| `description` | When Claude should delegate to this agent |
+| `tools` | Allowed tools (Grep, Glob, Read, WebSearch, Bash, Write, Edit) |
+| `model` | `haiku` for fast/cheap, `sonnet` for complex |
+
+---
+
+## Output Structure (Required)
+
+All agents MUST return structured findings:
+
+| Section | Required | Content |
+|---------|----------|---------|
+| RESULT | Yes | `found` / `not-found` / `partial` |
+| FILES FOUND | If found | File paths with line numbers |
+| PATTERNS | If found | Existing code patterns |
+| GAPS | If any | What's missing |
+| RECOMMENDATION | Yes | Specific next action |
+
+---
+
+## Best Practices
+
+- Write clear `description` - Claude uses this to decide when to delegate
+- Always include file:line references
+- Be specific in recommendations
+- Use `haiku` model for simple searches
+- Use `sonnet` for complex reasoning

package/templates/settings.json

@@ -0,0 +1,47 @@
+{
+  "permissions": {
+    "allow": [
+      "Bash(git status:*)",
+      "Bash(git diff:*)",
+      "Bash(git log:*)",
+      "Bash(git branch:*)",
+      "Bash(git add:*)",
+      "Bash(git commit:*)",
+      "Bash(git ls-tree:*)",
+      "Bash(git ls-files:*)",
+      "Bash(npm run:*)",
+      "Bash(npm test:*)",
+      "Bash(npm run build:*)",
+      "Bash(npm run lint:*)"
+    ]
+  },
+  "hooks": {
+    "SessionStart": [
+      {
+        "matcher": "startup",
+        "hooks": [
+          {
+            "type": "command",
+            "command": ".claude/hooks/session-start.sh",
+            "statusMessage": "Loading project context..."
+          }
+        ]
+      }
+    ],
+    "UserPromptSubmit": [
+      {
+        "matcher": ".*",
+        "hooks": [
+          {
+            "type": "command",
+            "command": ".claude/hooks/smart-inject-llm.sh",
+            "statusMessage": "Analyzing prompt..."
+          }
+        ]
+      }
+    ]
+  },
+  "sandbox": {
+    "enabled": true
+  }
+}