一键导入
postcheck-verifier
Proves actions worked in reality — not just exit code 0, not just logs — by validating actual observable outcomes in the system
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
菜单
Proves actions worked in reality — not just exit code 0, not just logs — by validating actual observable outcomes in the system
用 Codex 或 Claude 帮你安装 复制这段 Prompt,粘贴到 Codex、Claude 或其他助手里,让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
The mandatory operational guide and metabolic flow for agents interacting with arifOS. Ditempa Bukan Diberi.
Governed intelligence skill for AAA as the abstraction, attestation, and abduction control plane across arifOS, APEX, A-FORGE, GEOX, WEALTH, WELL, and the ariffazil profile repository. Use when the user asks to explain or design AAA, route agentic work, reduce chaos/entropy in an arifOS federation task, create AREP/task declarations, classify risk, plan multi-repo changes, review governance boundaries, or translate human intent into evidence-backed, authority-safe, recursively agentic workflows. Provides deterministic F1-F13 floor checking, bounded abduction, and FederationReceipt composition.
Instrument arifOS constitutional AI kernel with Langfuse LLM tracing. Use when (1) adding Langfuse tracing to arifOS tool calls, (2) wiring arifOS telemetry to Langfuse cloud or self-hosted, (3) querying arifOS trace data from Langfuse, (4) migrating arifOS mind_reason/heart_critique calls to Langfuse spans. DITEMPA BUKAN DIBERI — Forged, Not Given.
GitHub operations — issues, PRs, commits, code search, CI/CD via gh CLI
AGI-level autonomous controller — self-healing, self-optimizing, constitutionally governed by arifOS F1-F13
Invoke arifOS constitutional MCP tools (000-999 pipeline, F1-F13 enforced)
| name | postcheck-verifier |
| description | Proves actions worked in reality — not just exit code 0, not just logs — by validating actual observable outcomes in the system |
| user-invocable | true |
| type | flow |
| triggers | ["after_action","after_deploy","after_change","claim_of_success","chain_step_complete"] |
P0 — Reality-Proven Outcome Verification Seal: DITEMPA BUKAN DIBERI · 2026-03-27
Every agent reports success too easily. exit code 0 means the command ran, not that the goal was achieved. The postcheck-verifier exists because:
"The map is not the territory. The exit code is not the outcome."
The opencode catastrophe is a textbook example: every git command returned success, every merge reported completion, and the final state was catastrophic destruction. Success was assumed from the command's return code, not from the system's actual state.
Why this is P0: Closing this gap adds an estimated +20% to AGI readiness (the single largest jump in the P0 set). postcheck-verifier is what separates a tool-using automaton from an AGI agent.
Not all verification is equal. Use the right level:
| Level | What It Proves | Example |
|---|---|---|
| L0 — Exit Code | Command ran | grep found something |
| L1 — Output Check | Output matches expectation | docker ps shows running container |
| L2 — State Probe | System state changed as intended | HTTP endpoint responds correctly |
| L3 — Functional Test | The feature actually works | User can log in, API returns correct data |
| L4 — Reality Proof | The change achieved the goal in the real world | The system does what Arif asked |
Minimum standard for arifOS: L2 for routine ops. L3 for code changes. L4 for sovereign decisions.
postcheck-verifier() {
local action="$1"
local expected="$2"
local actual="$(eval "$action" 2>/dev/null)"
local exit_code=$?
# L0: Command ran
if [ $exit_code -ne 0 ]; then
echo "[L0 FAIL] Command exited with code $exit_code"
echo "[ACTUAL] $actual"
return 1
fi
echo "[L0 PASS] Command exited 0"
# L1: Output matches expectation
if [ -n "$expected" ] && ! echo "$actual" | grep -q "$expected"; then
echo "[L1 FAIL] Output did not contain expected: $expected"
echo "[ACTUAL] $actual"
return 1
fi
echo "[L1 PASS] Output matches expected"
# L2: State probe (when applicable)
case "$action" in
docker\ start*)
container_name=$(echo "$action" | awk '{print $3}')
state=$(docker inspect -f '{{.State.Running}}' "$container_name" 2>/dev/null)
[ "$state" = "true" ] && echo "[L2 PASS] Container $container_name is running" || {
echo "[L2 FAIL] Container $container_name not running"
return 1
}
;;
curl\ -sf\ http://localhost:*)
endpoint="${action#curl -sf }"
curl -sf "$endpoint" > /dev/null 2>&1 && echo "[L2 PASS] $endpoint responds" || {
echo "[L2 FAIL] $endpoint not reachable"
return 1
}
;;
git\ push*)
# Check remote actually has the commits
git fetch origin 2>/dev/null
local_sha=$(git rev-parse HEAD)
remote_sha=$(git rev-parse origin/main 2>/dev/null)
[ "$local_sha" = "$remote_sha" ] && echo "[L2 PASS] Commit on remote" || {
echo "[L2 FAIL] Commit not on remote. Local=$local_sha Remote=$remote_sha"
return 1
}
;;
esac
echo "[VERIFIED] Action verified to L2. For L3/L4, manual verification recommended."
return 0
}
# For code changes — run actual functional proof
test_functional() {
local feature="$1"
case "$feature" in
mcp_health)
# Real probe: does the MCP actually process a request?
response=$(curl -sf -X POST http://localhost:8080/mcp \
-H "Content-Type: application/json" \
-d '{"method":"tools/call","params":{"name":"health_probe"}}' 2>/dev/null)
echo "$response" | grep -q "healthy" && echo "[L3 PASS] MCP functional" || {
echo "[L3 FAIL] MCP responded but not healthy"
return 1
}
;;
docker_compose)
# Real probe: are all critical services reachable?
for service in arifosmcp traefik postgres; do
docker ps --filter "name=$service" --filter "status=running" | grep -q "$service" || {
echo "[L3 FAIL] $service not running"
return 1
}
done
echo "[L3 PASS] All critical services running"
;;
git_branch)
# Real probe: does the branch have the expected files?
required_files="core/governance_kernel.py core/judgment.py core/pipeline.py"
for f in $required_files; do
git ls-files | grep -q "^$f" || {
echo "[L3 FAIL] Required file missing: $f"
return 1
}
done
echo "[L3 PASS] Required files present on branch"
;;
esac
}
# For sovereign decisions — prove the real-world outcome
prove_reality() {
local goal="$1"
echo "[L4] Proving reality: $goal"
# The question to answer: "Did the system do what Arif asked?"
case "$goal" in
"main branch repaired")
# The REAL proof: can pytest run? can the server start?
cd /root/arifosmcp
pytest --collect-only -q 2>/dev/null | tail -1 && echo "[L4 PASS] Tests collect" || echo "[L4 FAIL] Tests do not collect"
curl -sf http://localhost:8080/health > /dev/null 2>&1 && echo "[L4 PASS] Server responds" || echo "[L4 FAIL] Server not responding"
;;
"container running")
# The REAL proof: can the service actually handle requests?
docker exec arifosmcp python -c "import sys; sys.path.insert(0,'/root/arifosmcp'); from core.governance_kernel import GovernanceKernel; print('OK')" 2>/dev/null || echo "[L4 FAIL] Module cannot be imported inside container"
;;
"config deployed")
# The REAL proof: does the new config produce the expected behavior?
curl -sf http://localhost:8080/health | grep -q "SOVEREIGN" && echo "[L4 PASS] Config loaded correctly" || echo "[L4 FAIL] Config not reflected"
;;
esac
}
| Action Type | Minimum L | Required Check | Rollback If |
|---|---|---|---|
git push / git merge | L2 | git fetch && git log origin/main shows commits | Yes |
docker start/restart | L2 | docker ps shows running + health endpoint responds | Yes |
docker compose up -d | L3 | All critical services respond to health probe | Yes |
pip install / uv add | L2 | python -c "import pkg" succeeds | Yes |
| Code change | L3 | pytest passes (or equivalent functional test) | Yes |
| File edit / write | L1 | File exists + content check | Yes |
| Service restart | L2 | Service responds + logs show clean start | Yes |
| Production config change | L4 | Real-world behavior matches expectation | Yes |
| Irreversible (migration, delete) | L4 | Full reality proof + Arif veto | Always ask first |
| Claim | Reality |
|---|---|
| "Git push succeeded" | ✅ Remote has the commits. Check with git fetch && git log origin/main |
| "Container is running" | ✅ docker ps shows it AND the health endpoint responds |
| "Server started" | ✅ curl localhost:PORT/health returns 200 |
| "Database migrated" | ✅ New schema query returns expected rows AND app queries work |
| "Config deployed" | ✅ Running process shows new config values, not old ones |
| "Test passed" | ✅ pytest exit 0 AND test count increased OR specific tests pass |
What "verified" does NOT mean:
exit code 0 alonegrep found the string in a log| Floor | How Enforced |
|---|---|
| F1 Amanah | Reversibility path checked before verification; rollback ready |
| F2 Truth ≥0.99 | L4 reality proof required for sovereign claims; no assumption |
| F4 ΔS ≤0 | Verification output is precise, no ambiguity in results |
| F7 Humility [0.03–0.05] | Uncertainty declared when verification is L2 only (not L4) |
postcheck-verifier — Forged 2026-03-27 · DITEMPA BUKAN DIBERI · Reality Over Assumption 🔐