agents/_lanes/333-AGI/skills/agi_mcp_discover/SKILL.md
# Skill: agi_mcp_discover **Agent:** 333-AGI **Namespace:** agi_* **Cluster:** TOOL **Trigger:** Before loading MCP tools for a task **Capability:** Return only the tools needed for a given task domain. Always-hot tools (kernel verbs + health) + on-demand tools (geoscience/capital/engineering/vault/meta/evidence). Capped at 8 total to protect context window. **MCP tools underneath:** self (agi_mcp_discover is both caller and callee) **Blast radius:** LOW — read-only query **888 gate required:**
npx skillsauth add ariffazil/openclaw-workspace agents/_lanes/333-AGI/skills/agi_mcp_discoverInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Agent: 333-AGI Namespace: agi_* Cluster: TOOL Trigger: Before loading MCP tools for a task Capability: Return only the tools needed for a given task domain. Always-hot tools (kernel verbs + health) + on-demand tools (geoscience/capital/engineering/vault/meta/evidence). Capped at 8 total to protect context window. MCP tools underneath: self (agi_mcp_discover is both caller and callee) Blast radius: LOW — read-only query 888 gate required: No (read-only tool selection)
agi_mcp_discover is the progressive MCP tool loader for 333-AGI (Δ MIND).
The 333-AGI runtime normally sees ~49 A-FORGE tools + GEOX + WEALTH + WELL + arifOS in a single MCP context. Loading all of them before the model has even read the user request blows the context window and forces the agent to relearn the same surface on every task.
This skill enforces a hard cap of 8 tools per MCP context (F4 CLARITY). Seven tools are always hot (kernel verbs + health + vitality + self). One domain — selected by classify_intent() — gets the rest of the room.
from agi_mcp_discover import discover, classify_intent
# What domain does this user want?
domain = classify_intent("Compute NPV for the Malay CCS deal")
# → "capital"
# Load only the tools we actually need.
tools = discover(domain)
# → 7 always-hot + 6 on-demand = 8 (capped)
| Symbol | Type | Purpose |
|-------------------|-------------------------------|------------------------------------------------------|
| ALWAYS_HOT | list[str] | 7 tools loaded on every task |
| ON_DEMAND | dict[str, list[str]] | Domain → on-demand tool set |
| MAX_TOOLS | int = 8 | Hard cap (F4 CLARITY) |
| discover(domain) | (str, str \| None) → list[str] | Domain-keyed tool loader, capped at MAX_TOOLS |
| classify_intent(s) | (str) → str | Keyword router: user intent → task domain |
classify_intent() matches the first hit. Order matters:
geoscience — basin, geology, seismic, petrophysics, well, prospectcapital — npv, irr, emv, capital, invest, valuation, stockengineering — code, build, deploy, shell, docker, git, testvault — vault, seal, ledger, archiveevidence — search, research, findmeta — fallback (agent_search, skill_load, agent_status)| Tool | Purpose |
|----------------------------|----------------------------------------------|
| arif_init | Bind cognitive runtime to active session |
| arif_observe | OBSERVE-stage verb |
| arif_think | THINK-stage verb |
| forge_probe | Federation organ liveness |
| forge_status | Jobs/leases/agents overview |
| well_assess_homeostasis | Vitality gate (Floor compliance) |
| agi_mcp_discover | Self-reference (this skill) |
| Domain | Tools added |
|---------------|-------------------------------------------------------------------------------|
| geoscience | 7 GEOX tools (basin, seismic, petrophysics, prospect, deep time, well ingest) |
| capital | 6 WEALTH tools (npv, irr, emv, runway, stock, diagnose) |
| engineering | 6 A-FORGE tools (shell, git, docker, fs read/write, health check) |
| vault | 3 vault tools + forge_vault_seal |
| meta | 3 (agent_search, skill_load, agent_status) |
| evidence | 3 (forge_research, forge_minimax_search, forge_docs_lookup) |
meta.PYTHONPATH=/root/AAA/agents/_lanes/333-AGI/runtime \
python3 -m unittest discover -s /root/AAA/agents/_lanes/333-AGI/runtime/tests -v
Coverage: domain classification, hard cap, always-hot invariants, no-duplicate keys, intent param forward-compat.
agi_tool_invoke (already in agent-card) owns that.planner/ (sibling skill) owns DAG construction.Phase 4b · AAA entropy-reduction roadmap · F13 SOVEREIGN.
testing
OpenClaw edge agent bridge — operational triage, doctor, restart, and A2A bridge routing for the federation edge (Telegram surface). USE WHEN: "openclaw unhealthy", "gateway down", "edge bot not responding", "a2a bridge disconnected", "watchdog tripped", "openclaw doctor", "openclaw restart". NOT for token/security audit — use FORGE-telegram-audit.
tools
Generate images, videos, TTS, voice clone, and music via MiniMax MCP server. Use when user asks to "draw", "generate image", "create picture", "make a photo", "text to image", "image generation".
testing
Single load-bearing constitutional-judgment skill. Routes all F1–F13, verdict, hold, seal, scope, authority and floor-check calls through the live arif_judge surface. Replaces 7 overlapping predecessors (arifos-constitutional-judge, arifos-constitutional-judge, arifos-constitutional-judge, arifos-constitutional-judge, arifos-constitutional-judge, arifos-constitutional-judge, arifos-constitutional-judge).
development
MANDATORY LSP grounding gate BEFORE any code mutation on .ts, .py, .js, .tsx, .jsx files. Forces the agent to read real-time compiler diagnostics and structural project context before editing — eliminating blind guesses and anchoring every mutation in F2 (TRUTH). Routes through arifOS kernel (:8088) for centralized gate logic.