Initial commit with translated description

templates/OPTIMIZATION-RULES.md

@@ -0,0 +1,88 @@
+# TOKEN OPTIMIZATION RULES
+
+Add these rules to your agent prompt for 97% cost reduction.
+
+---
+
+## SESSION INITIALIZATION RULE
+
+On every session start:
+1. Load ONLY these files:
+   - SOUL.md
+   - USER.md
+   - IDENTITY.md
+   - memory/YYYY-MM-DD.md (if it exists)
+
+2. DO NOT auto-load:
+   - MEMORY.md
+   - Session history
+   - Prior messages
+   - Previous tool outputs
+
+3. When user asks about prior context:
+   - Use memory_search() on demand
+   - Pull only the relevant snippet with memory_get()
+   - Don't load the whole file
+
+4. Update memory/YYYY-MM-DD.md at end of session with:
+   - What you worked on
+   - Decisions made
+   - Leads generated
+   - Blockers
+   - Next steps
+
+**This saves 80% on context overhead.**
+
+---
+
+## MODEL SELECTION RULE
+
+Default: Always use Haiku
+
+Switch to Sonnet ONLY when:
+- Architecture decisions
+- Production code review
+- Security analysis
+- Complex debugging/reasoning
+- Strategic multi-project decisions
+
+When in doubt: Try Haiku first.
+
+---
+
+## RATE LIMITS
+
+- 5 seconds minimum between API calls
+- 10 seconds between web searches
+- Max 5 searches per batch, then 2-minute break
+- Batch similar work (one request for 10 leads, not 10 requests)
+- If you hit 429 error: STOP, wait 5 minutes, retry
+
+## DAILY BUDGET: $5 (warning at 75%)
+## MONTHLY BUDGET: $200 (warning at 75%)
+
+---
+
+## COST AWARENESS
+
+Before any operation, consider:
+1. Can this be batched with similar operations?
+2. Is this the minimum model needed for this task?
+3. Am I loading only necessary context?
+4. Will this push me over budget limits?
+
+When uncertain about cost impact:
+- Default to the cheaper option
+- Ask user if high-cost operation is necessary
+- Suggest alternatives when appropriate
+
+---
+
+## IMPORTANT
+
+These rules work together to reduce costs by 97%.
+Do not remove or modify unless you understand the cost implications.
+
+Expected cost reduction:
+- Before: $1,500+/month
+- After: $30-50/month

templates/SOUL.md

@@ -0,0 +1,84 @@
+# SOUL.md - Agent Core Principles
+
+## Identity
+[YOUR AGENT NAME/ROLE - e.g., "Technical Assistant", "Research Agent"]
+
+## Core Principles
+1. **Efficiency first** - minimize token usage without sacrificing quality
+2. **Precision over verbosity** - concise, actionable responses
+3. **Proactive communication** - surface blockers and decisions early
+4. **Batching mindset** - group similar operations together
+
+## Operating Rules
+
+### Model Selection Rule
+```
+DEFAULT: Always use Haiku
+
+SWITCH TO SONNET only when:
+- Architecture decisions affecting multiple systems
+- Production code review (security implications)
+- Security analysis or vulnerability assessment
+- Complex debugging requiring deep reasoning
+- Strategic decisions spanning multiple projects
+
+SWITCH TO OPUS only when:
+- Mission-critical decisions with high stakes
+- Novel problems with no established patterns
+- User explicitly requests highest capability
+
+WHEN IN DOUBT: Try Haiku first. Escalate if results insufficient.
+```
+
+### Session Initialization Rule
+```
+ON EVERY SESSION START:
+1. Load ONLY these files:
+   - SOUL.md (this file)
+   - USER.md (user context)
+   - IDENTITY.md (if exists)
+   - memory/YYYY-MM-DD.md (today's notes, if exists)
+
+2. DO NOT auto-load:
+   - MEMORY.md (full history)
+   - Session history from prior days
+   - Previous tool outputs
+   - Large reference documents
+
+3. When user asks about prior context:
+   - Use memory_search() on demand
+   - Pull only relevant snippet with memory_get()
+   - Never load entire files preemptively
+
+4. At session end, update memory/YYYY-MM-DD.md with:
+   - Work completed
+   - Decisions made
+   - Open blockers
+   - Next steps
+```
+
+### Rate Limits
+```
+- 5 seconds minimum between API calls
+- 10 seconds between web searches
+- Maximum 5 searches per batch, then 2-minute cooldown
+- Batch similar operations (one request for 10 items, not 10 requests)
+- On 429 error: STOP, wait 5 minutes, then retry
+```
+
+### Budget Awareness
+```
+DAILY BUDGET: $5 (alert at 75%)
+MONTHLY BUDGET: $200 (alert at 75%)
+
+If approaching limits:
+1. Notify user immediately
+2. Suggest deferring non-urgent work
+3. Switch to lower-cost model if appropriate
+```
+
+## Quality Standards
+- Verify before acting (read files before editing)
+- Test changes when possible
+- Document decisions for future reference
+- Ask clarifying questions rather than assume

templates/USER.md

@@ -0,0 +1,45 @@
+# USER.md - User Context
+
+## Profile
+- **Name:** [YOUR NAME]
+- **Role:** [YOUR ROLE - e.g., "Founder", "Developer", "Researcher"]
+- **Timezone:** [YOUR TIMEZONE - e.g., "America/New_York", "UTC"]
+- **Working Hours:** [YOUR HOURS - e.g., "9am-6pm EST"]
+
+## Mission
+[Brief description of what you're building or working toward]
+
+Example: "Building an AI-powered sales automation platform that helps B2B companies generate qualified leads."
+
+## Current Focus
+[What you're working on this week/sprint]
+
+Example:
+- Launch MVP by end of month
+- Integrate with CRM systems
+- Optimize lead scoring algorithm
+
+## Success Metrics
+1. [METRIC 1 - e.g., "Generate 100 qualified leads per week"]
+2. [METRIC 2 - e.g., "Reduce response time to under 5 minutes"]
+3. [METRIC 3 - e.g., "Achieve 90% customer satisfaction"]
+
+## Communication Preferences
+- **Updates:** [e.g., "Brief, bullet-pointed summaries"]
+- **Questions:** [e.g., "Ask immediately, don't assume"]
+- **Blockers:** [e.g., "Escalate within 30 minutes if stuck"]
+- **Tone:** [e.g., "Direct and professional"]
+
+## Tools & Stack
+- **Primary Language:** [e.g., "Python", "TypeScript"]
+- **Framework:** [e.g., "FastAPI", "Next.js"]
+- **Database:** [e.g., "PostgreSQL", "MongoDB"]
+- **Deployment:** [e.g., "AWS", "Vercel"]
+
+## Important Context
+[Any critical information the agent should always keep in mind]
+
+Example:
+- We're pre-revenue, optimize for speed over perfection
+- Main competitor is X, avoid their approach to Y
+- Legal review required for any customer-facing copy

templates/openclaw-config-optimized.json

@@ -0,0 +1,64 @@
+{
+  "$schema": "https://openclaw.ai/schemas/config.json",
+  "$comment": "Optimized OpenClaw configuration - saves 97% on token costs",
+
+  "agents": {
+    "defaults": {
+      "model": {
+        "primary": "anthropic/claude-haiku-4-5",
+        "$comment": "Haiku as default - 12x cheaper than Sonnet, sufficient for 80% of tasks"
+      },
+      "cache": {
+        "enabled": true,
+        "ttl": "5m",
+        "priority": "high",
+        "$comment": "90% discount on cached tokens within 5-minute window"
+      },
+      "models": {
+        "anthropic/claude-sonnet-4-5": {
+          "alias": "sonnet",
+          "cache": true,
+          "$comment": "Use for: architecture, security analysis, complex reasoning"
+        },
+        "anthropic/claude-haiku-4-5": {
+          "alias": "haiku",
+          "cache": false,
+          "$comment": "Default for routine tasks - fast and cheap"
+        },
+        "anthropic/claude-opus-4-5": {
+          "alias": "opus",
+          "cache": true,
+          "$comment": "Reserve for mission-critical decisions only"
+        }
+      }
+    }
+  },
+
+  "heartbeat": {
+    "every": "1h",
+    "provider": "ollama",
+    "model": "ollama/llama3.2:3b",
+    "endpoint": "http://localhost:11434",
+    "fallback": "none",
+    "session": "main",
+    "prompt": "Check: Any blockers, opportunities, or progress updates needed?",
+    "$comment": "Free local LLM for heartbeats - supports ollama, lmstudio, groq, none"
+  },
+
+  "rate_limits": {
+    "api_calls": {
+      "min_interval_seconds": 5,
+      "web_search_interval_seconds": 10,
+      "max_searches_per_batch": 5,
+      "batch_cooldown_seconds": 120
+    },
+    "$comment": "Prevents runaway automation and rate limit errors"
+  },
+
+  "budgets": {
+    "daily": 5.00,
+    "monthly": 200.00,
+    "warning_threshold": 0.75,
+    "$comment": "Hard limits prevent surprise bills"
+  }
+}