Agent skill
arif-init-wrapper
Universal session auto-anchor for ALL agent vendors on arifOS VPS — Codex, Claude, Gemini, Kimi, OpenCode, Aider, Copilot, and future agents
Install this agent skill to your Project
npx add-skill https://github.com/ariffazil/arifOS/tree/main/skills/arif-init-wrapper
SKILL.md
arif-init-wrapper
P0 — Universal Anchor for All Agent Vendors Seal: DITEMPA BUKAN DIBERI · 2026-03-27
Purpose
Every agent on this VPS starts from the same broken state: no anchor, no session context, no constitutional grounding. The arif-init-wrapper closes this gap for every vendor by providing a universal bootstrap that:
- Establishes a governed session via arifOS MCP
- Verifies the agent's identity context
- Checks constitutional floor health before any tool is touched
- Reads carry-forward context from memory before starting work
- Reports its status to the operator in plain language
Why this is P0: Without this, agents operate unanchored — the exact condition that produced the opencode catastrophe (session with no constitutional grounding, leading to blind git reset --hard origin/main).
Trigger Patterns
Every time any of these agents start a session, run arif-init-wrapper:
| Agent | Invocation |
|---|---|
| Claude Code | On every claude session start |
| Codex | On every codex session start |
| Gemini CLI | On every gemini session start |
| Kimi | On every kimi session start |
| OpenCode | On every opencode session start |
| Aider | On every aider session start |
| Copilot CLI | On every github-copilot-cli session start |
| Any new agent | On first boot, before any tool use |
Protocol
Step 1 — Anchor Session
# Probe arifOS MCP availability
curl -sf http://localhost:8080/health 2>/dev/null && {
# Session anchor via MCP
arifos_mcp_call tool=init_anchor payload='{"actor_id":"AGENT","session_purpose":"WORK"}'
} || {
echo "[UNANCHORED] arifOS MCP unavailable. Operating WITHOUT constitutional grounding."
echo "[WARNING] This is high-risk. Proceed with extreme caution."
}
Step 2 — Constitutional Integrity Check (LAW 2)
Before any work, verify the foundation:
# Check core directory integrity
core_modules=$(ls /root/arifosmcp/core/*.py 2>/dev/null | wc -l)
if [ "$core_modules" -lt 10 ]; then
echo "[P0 CONSTITUTIONAL CRISIS] /root/arifosmcp/core/ appears broken ($core_modules modules, expected 20+)"
echo "[ACTION] HALT. Report to Arif. Do not proceed."
echo "[RECOVERY] Check feature branches: git branch -a | grep feature/"
return 1 2>/dev/null || exit 1
fi
# Check pytest can collect tests (structural sanity)
cd /root/arifosmcp && pytest --collect-only -q 2>/dev/null | tail -1 && {
echo "[OK] Constitutional structure verified."
} || {
echo "[WARNING] pytest collection failed. Foundation may be unstable."
}
Step 3 — Check Floors (F1–F13 baseline)
# Verify critical floors are loadable
python3 -c "
import sys
sys.path.insert(0, '/root/arifosmcp')
try:
from core.shared.floors import THRESHOLDS
floors_ok = len(THRESHOLDS) >= 13
print(f'[FLOORS] {len(THRESHOLDS)}/13 loaded')
except Exception as e:
print(f'[FATAL] Floor check failed: {e}')
sys.exit(1)
"
Step 4 — Memory Carry-Forward
# Query recent scars and unresolved issues
arifos_mcp_call tool=memory_query payload='{"query":"recent scars unresolved actions","limit":5}'
Step 5 — Agent Identity Declaration
State clearly:
- Which agent vendor is running
- Whether constitutional grounding is active or unavailable
- Any constitutional warnings for this session
Step 6 — Operator Report
Report to Arif in plain language:
[ANCHOR] Agent: {vendor}
[STATUS] arifOS MCP: {available/unavailable}
[FLOORS] Constitutional check: {pass/wail}
[CORE] Core modules: {count}/expected
[MEMORY] Recent context: {carry_forward_summary}
[RISK] Session risk level: {LOW/MEDIUM/HIGH/CRITICAL}
[ACTION] Ready to proceed: YES/NO
Decision Matrix
| arifOS MCP | Core OK | pytest OK | Floors OK | Action |
|---|---|---|---|---|
| ✅ | ✅ | ✅ | ✅ | Proceed normally |
| ✅ | ✅ | ✅ | ⚠️ | Proceed with floor warnings |
| ✅ | ⚠️ | ✅/⚠️ | ✅ | HALT — P0 Foundational Crisis |
| ✅ | ❌ | ❌ | ❌ | HALT — full corruption |
| ❌ | ✅ | ✅ | ✅ | HIGH RISK — proceed with extreme caution, log everything |
| ❌ | ⚠️/❌ | ⚠️/❌ | ⚠️/❌ | CRITICAL — HALT, report to Arif immediately |
Anti-Patterns This Closes
- Unanchored Codex —
codexruns without session context, blunders intogit reset --hard - Ungrounded Gemini —
geminiassumesorigin/mainis good without checking - Skip-before-check — Any agent that runs
git pullbefore verifying the repo is structurally sound - Silent failure — Agent operates in a broken state without telling Arif
Constitutional Alignment
| Floor | How Enforced |
|---|---|
| F1 Amanah | Reversibility checked before any destructive git op |
| F2 Truth ≥0.99 | Uncertainty bands declared in operator report |
| F7 Humility [0.03–0.05] | Explicit confidence level in session status |
| F11 CommandAuth | Protected identities require cryptographic proof |
| F13 Sovereign | All HALT states require Arif's explicit decision to proceed |
arif-init-wrapper — Forged 2026-03-27 · DITEMPA BUKAN DIBERI · Universal Agent Anchor 🔐
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
mcp-config-separation
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.
MCP_CONFIG
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.
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.
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.
Didn't find tool you were looking for?