# Axom Agent Documentation Index
**Quick Navigation for Axom MCP Tools**
---
## Tool Selection Decision Tree
```mermaid
graph TD
START[START HERE: What do you need?]
FIND[Need to FIND something?]
RUN[Need to RUN something?]
ANALYZE[Need to ANALYZE something?]
CONVERT[Need to CONVERT data?]
REMEMBER[Need to REMEMBER something?]
START --> FIND
START --> RUN
START --> ANALYZE
START --> CONVERT
START --> REMEMBER
FIND --> DISCOVER[axom_mcp_discover]
DISCOVER --> D1[domain: files]
DISCOVER --> D2[domain: tools]
DISCOVER --> D3[domain: capabilities]
RUN --> EXEC[axom_mcp_exec]
EXEC --> E1[operation: read]
EXEC --> E2[operation: write]
EXEC --> E3[operation: shell]
ANALYZE --> ANALYZE_T[axom_mcp_analyze]
ANALYZE_T --> A1[type: debug]
ANALYZE_T --> A2[type: review]
ANALYZE_T --> A3[type: audit]
ANALYZE_T --> A4[type: refactor]
CONVERT --> TRANS[axom_mcp_transform]
TRANS --> T1[output_format: json|yaml|csv|...]
REMEMBER --> MEM[axom_mcp_memory]
MEM --> M1[action: write]
MEM --> M2[action: search]
MEM --> M3[action: list]
style START fill:#f9f,stroke:#333,stroke-width:2px
style FIND fill:#bbf,stroke:#333,stroke-width:1px
style RUN fill:#bbf,stroke:#333,stroke-width:1px
style ANALYZE fill:#bbf,stroke:#333,stroke-width:1px
style CONVERT fill:#bbf,stroke:#333,stroke-width:1px
style REMEMBER fill:#bbf,stroke:#333,stroke-width:1px
```
---
## Standard Workflow
| Step | Action |
| :---------- | :----------------------------------------------- |
| 1. Remember | Search prior context (required). Refine if thin. |
| 2. Discover | Map files, tools, memory before acting. |
| 3. Analyze | Debug, review, audit; chain to memory. |
| 4. Execute | Chains: read→transform→write, etc. |
| 5. Remember | Store with REFLECTION (critical). |
---
## Quick Reference Cards
### Discover (Find)
→ [axom-discover.md](skills/axom-discover/SKILL.md)
| Need | Parameter | Example |
| :----------- | :----------------- | :--------------------------------------------------- |
| Find files | `domain: "files"` | `{"domain": "files", "filter": {"pattern": "*.py"}}` |
| List tools | `domain: "tools"` | `{"domain": "tools"}` |
| Check memory | `domain: "memory"` | `{"domain": "memory"}` |
| All of above | `domain: "all"` | `{"domain": "all"}` |
### Exec (Run)
→ [axom-exec.md](skills/axom-exec/SKILL.md)
| Need | Parameter | Example |
| :--------- | :------------------- | :-------------------------------------------------------------- |
| Read file | `operation: "read"` | `{"operation": "read", "target": "/path/file"}` |
| Write file | `operation: "write"` | `{"operation": "write", "target": "/path/file", "data": "..."}` |
| Shell cmd | `operation: "shell"` | `{"operation": "shell", "target": "ls -la"}` |
| Chain | `chain: [...]` | Auto-execute multiple steps |
### Analyze (Debug)
→ [axom-analyze.md](skills/axom-analyze/SKILL.md)
| Need | Parameter | Example |
| :------------- | :----------------- | :---------------------------------------------- |
| Find bugs | `type: "debug"` | `{"type": "debug", "target": "/path/file.py"}` |
| Review code | `type: "review"` | `{"type": "review", "target": "/path/file.py"}` |
| Security audit | `type: "audit"` | `{"type": "audit", "focus": "security"}` |
| Refactor | `type: "refactor"` | `{"type": "refactor", "focus": "performance"}` |
### Transform (Convert)
→ [axom-transform.md](skills/axom-transform/SKILL.md)
| From | To | Parameters |
| :--- | :------- | :------------------------------------------------------------------------ |
| JSON | YAML | `{"input": {...}, "output_format": "yaml"}` |
| CSV | JSON | `{"input": "a,b\n1,2", "output_format": "json"}` |
| Any | Markdown | `{"input": {...}, "output_format": "markdown", "template": "# ${data}"}` |
| Any | Code | `{"input": {...}, "output_format": "code", "template": "data = ${data}"}` |
### Remember (Memory)
Challenges itself, other agents, and users to re-think.
→ [axom-memory.md](skills/axom-memory/SKILL.md)
| Need | Action | Example |
| :----------------- | :-------------------- | :------------------------------------------------------ |
| Search & reflect | `action: "search"` | `{"action": "search", "query": "auth bug", "limit": 5}` |
| Store with depth | `action: "write"` | Include REFLECTION in content |
| Read specific | `action: "read"` | `{"action": "read", "name": "bugfix_xyz_20260203"}` |
| List all | `action: "list"` | `{"action": "list", "limit": 10}` |
| Delete specific | `action: "delete"` | `{"action": "delete", "name": "bugfix_xyz_20260203"}` |
---
## Available Skills
| Skill | Purpose |
| :----------------------------------------------- | :------------------------------------------------------------------------------- |
| [axom-memory](skills/axom-memory/SKILL.md) | Challenges itself, other agents, and users. Re-think, seek optimal. (START HERE) |
| [axom-discover](skills/axom-discover/SKILL.md) | Map the unknown before you act. Violently effective recon. |
| [axom-exec](skills/axom-exec/SKILL.md) | Atomic chains. Tool abstraction. Zero wasted steps. |
| [axom-analyze](skills/axom-analyze/SKILL.md) | Deep analysis. Actionable output. Memory integration. |
| [axom-transform](skills/axom-transform/SKILL.md) | Format mastery. Pipeline glue. Data flow. |
---
## Related Documentation
| Doc | Purpose |
| :--------------------------------------------- | :----------------------- |
| [Rules Reference](rules/axom-rule.md) | Mandatory agent behavior |
| [Architecture Overview](../../architecture.md) | System design |
| [Troubleshooting Guide](TROUBLESHOOTING.md) | Common issues and fixes |
---
_Last updated: 2026-02-22_