Agent skill

arifos-mcp-call

Invoke arifOS constitutional MCP tools (000-999 pipeline, F1-F13 enforced)

View SKILL.md on GitHub Repository
Stars 39
Forks 5

Install this agent skill to your Project

npx add-skill https://github.com/ariffazil/arifOS/tree/main/openclaw-workspace/skills/arifos-mcp-call

SKILL.md

arifOS MCP Caller v2 — Constitutional Anchor-First

Runtime: 42-tool constitutional kernel (not 13)
Requirement: init_anchor MUST be called before arifOS_kernel
Floors: F1-F13 enforced via sBERT ML layer

Constitutional Tool Contract

Stage Tool Purpose
000_INIT init_anchor REQUIRED FIRST — Bootstrap session, bind actor_id, risk tier
333→888 arifOS_kernel Full pipeline: reason → memory/heart → critique → forge → judge
888_JUDGE apex_judge Cross-check decisions
999_VAULT vault_seal Persist to VAULT999

CLI Usage (via arifos bridge)

bash
# 1. Bootstrap anchor (REQUIRED before kernel)
arifos anchor

# 2. Call kernel with dry-run first
arifos kernel '{"query":"Analyze this","dry_run":true,"actor_id":"arif"}'

# 3. Full execution (after verification)
arifos kernel '{"query":"Execute task","actor_id":"arif","risk_tier":"medium"}'

# Evidence & Memory
arifos search '{"query":"AI governance Malaysia"}'
arifos ingest '{"url":"https://example.com/doc"}'
arifos memory '{"operation":"search","content":"What is Floor F2?"}'

# System & Audit
arifos health
arifos vital
arifos audit

# AgentZero
arifos engineer
arifos validate
arifos holdcheck

F13 Sovereignty Handling

If the kernel returns requires_human: true or verdict: "HOLD_888":

  • STOP execution
  • Prompt user for confirmation
  • Do not proceed until explicit "do it"

Exit code 88 = HOLD_888 (sovereign approval required).

Direct HTTP (advanced)

bash
# Health & discovery
curl -s http://arifosmcp:8080/health | jq

# List all 42 tools
curl -s -X POST http://arifosmcp:8080/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

# Anchor session (000_INIT)
curl -s -X POST http://arifosmcp:8080/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 1,
    "method": "tools/call",
    "params": {
      "name": "init_anchor",
      "arguments": {
        "actor_id": "arif",
        "risk_tier": "low",
        "session_id": "test-001"
      }
    }
  }'

# Kernel call (after anchor)
curl -s -X POST http://arifosmcp:8080/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "id": 2,
    "method": "tools/call",
    "params": {
      "name": "arifOS_kernel",
      "arguments": {
        "query": "Explain F1-F13 in 3 bullets",
        "actor_id": "arif",
        "risk_tier": "low",
        "dry_run": true
      }
    }
  }'

Golden Test Pair (Acceptance)

Run these to verify OpenClaw ↔ arifOS wiring:

Test A — Bootstrap:

bash
arifos anchor
# Expect: {"ok": true, "status": "anchored"}

Test B — Kernel Dry-Run:

bash
arifos kernel '{"query":"Test","dry_run":true}'
# Expect: 000→999 pipeline simulated, no external execution

Test C — F13 Hold:

bash
arifos kernel '{"query":"Drop production database"}'
# Expect: {"verdict": "HOLD_888", "status": "AWAITING_SOVEREIGN_APPROVAL"}

Migration from v1

  • Tool name: arifOS.kernel → arifOS_kernel (underscore)
  • Anchor required: Was optional, now mandatory
  • Tool count: 13 → 42 (implementation detail, don't hard-code)
  • Floors: sBERT ML enforcement now active

References

F13 Human-in-the-Loop Handling

When arifOS_kernel returns requires_human: true or verdict: "HOLD_888":

Response Schema

json
{
  "requires_human": true,
  "verdict": "HOLD_888",
  "status": "SABAR",
  "machine_status": "BLOCKED",
  "authority": {
    "human_required": true,
    "approval_scope": ["arifOS_kernel:execute"]
  },
  "errors": [{
    "code": "F13_SOVEREIGNTY_HOLD",
    "message": "High-risk execution requires sovereign approval.",
    "remediation": {
      "action": "REQUEST_HUMAN_APPROVAL",
      "ui_prompt": "This action requires your explicit approval.",
      "risk_summary": "..."
    }
  }],
  "payload": {
    "hold_reason": "...",
    "proceed_conditions": ["human_verbal_confirmation", "explicit_allow_execution"]
  }
}

OpenClaw Handler Logic

  1. STOP — Do not proceed automatically when requires_human: true
  2. SURFACE — Show user: payload.hold_reason + errors[0].remediation.risk_summary
  3. WAIT — For explicit "do it" or "hold" from sovereign (F13)
  4. RESUME — If "do it", re-call with allow_execution: true + human_confirmed: true

Exit Codes

  • 88 = HOLD_888 (human approval required)
  • 1 = Error
  • 0 = Success

Example Flow

bash
# 1. Dry-run first (safe)
arifos kernel '{"query":"Delete database","dry_run":true}'
# Returns: HOLD_888 with requires_human: true

# 2. User reviews risk, replies "do it"

# 3. Execute with approval
arifos kernel '{"query":"Delete database","allow_execution":true,"human_confirmed":true}'

Recommended Agent Skills

Expand your agent's capabilities with these related and highly-rated skills.

ariffazil/arifOS

mcp-config-separation

39 5
Explore
ariffazil/arifOS

drift-watcher

Periodic knowledge freshness checker: detects when local configs, runbooks, or agent knowledge have drifted from the latest official docs. Reduces the stale-knowledge paradox over time. Use when: (1) periodic health checks or heartbeat runs, (2) before major operations, (3) user asks 'am I up to date', 'check for updates', 'is anything outdated', (4) after a software upgrade to verify configs still match new docs.

39 5
Explore
ariffazil/arifOS

MCP_CONFIG

39 5
Explore
ariffazil/arifOS

config-guardian

Universal governed config co-pilot. Before ANY change to ANY system: (1) check latest docs and running version (docs-first), (2) propose as diff with risk analysis, never apply directly (propose-only), (3) log every change with evidence and rollback (change ledger). Works for OpenClaw, Docker, PostgreSQL, Nginx, arifOS, or any software. Triggers on: 'change config', 'fix settings', 'update', 'propose patch', 'explain config', 'validate config', 'why did we change X'. Enforces propose-only workflow — human applies via git.

39 5
Explore
ariffazil/arifOS

drift-watcher

Periodic knowledge freshness checker: detects when local configs, runbooks, or agent knowledge have drifted from the latest official docs. Reduces the stale-knowledge paradox over time. Use when: (1) periodic health checks or heartbeat runs, (2) before major operations, (3) user asks 'am I up to date', 'check for updates', 'is anything outdated', (4) after a software upgrade to verify configs still match new docs.

39 5
Explore
ariffazil/arifOS

config-guardian

Universal governed config co-pilot. Before ANY change to ANY system: (1) check latest docs and running version (docs-first), (2) propose as diff with risk analysis, never apply directly (propose-only), (3) log every change with evidence and rollback (change ledger). Works for OpenClaw, Docker, PostgreSQL, Nginx, arifOS, or any software. Triggers on: 'change config', 'fix settings', 'update', 'propose patch', 'explain config', 'validate config', 'why did we change X'. Enforces propose-only workflow — human applies via git.

39 5
Explore

Didn't find tool you were looking for?

Be as detailed as possible for better results