/iblai-vibe-agent-audit

Mirrored from iblai/vibe · skills/agents/iblai-vibe-agent-audit/SKILL.md. This page is generated — edit it in the repository, not here.

Add the agent Audit tab -- a paginated audit log of agent-scoped changes (create / update / delete) with humanized action summaries (e.g. "Enabled tools on E2E Agent 123 (uuid)", "Updated forkable, LLM model, ... on Settings for E2E Agent"), filterable by actor email, date range, and action type. Renders an empty state, an error state, and a 403 state for users without audit-read permission.

Common setup (brand, conventions, env files, verification): see docs/skill-setup.md.

Prerequisites

• Auth must be set up first (/iblai-vibe-auth)
• MCP server + skills configured (@iblai/mcp in .mcp.json)
• Ask the user for a real mentorId (agent UUID). Do NOT invent one.
• The current user must have Ibl.Analytics/Core/read (viewer) or Ibl.* (organization admin) permission. Without it the component renders the 403 state — no tenantKey mismatch, just RBAC.

Step 1: Check Environment

Before proceeding, check for an iblai.env in the project root. Look for PLATFORM, DOMAIN, and TOKEN variables. If the file does not exist or is missing these variables, tell the user: "You need an iblai.env with your platform configuration. Download the template and fill in your values: curl -o iblai.env https://raw.githubusercontent.com/iblai/vibe/refs/heads/main/iblai.env"

Step 2: Mount AnalyticsAuditLogStats

AnalyticsAuditLogStats does not read from AgentSettingsProvider — it takes tenantKey, mentorId, and userId as required props. The host page is responsible for resolving them (typically from localStorage after auth).

// app/(app)/agents/[mentorId]/audit/page.tsx
"use client";

import { useEffect, useState } from "react";
import { useParams } from "next/navigation";
import { AnalyticsAuditLogStats } from "@iblai/iblai-js/web-containers";

export default function AgentAuditPage() {
  const { mentorId } = useParams<{ mentorId: string }>();
  const [tenantKey, setTenantKey] = useState("");
  const [userId, setUserId] = useState("");

  useEffect(() => {
    try {
      const raw = localStorage.getItem("userData");
      if (raw) {
        const parsed = JSON.parse(raw);
        setUserId(parsed.user_nicename ?? parsed.username ?? "");
      }

      const resolvedTenant =
        localStorage.getItem("app_tenant") ??
        (() => {
          try {
            return JSON.parse(localStorage.getItem("current_tenant") ?? "{}").key;
          } catch { return undefined; }
        })() ??
        localStorage.getItem("tenant") ??
        "";
      setTenantKey(resolvedTenant);
    } catch {}
  }, []);

  if (!tenantKey || !userId) return null;

  return (
    <div className="flex h-full flex-col bg-white p-6">
      <div className="mb-6">
        <h2 className="text-lg font-semibold text-gray-900">Audit</h2>
        <p className="text-sm text-gray-500">
          View who changed what and when for this agent.
        </p>
      </div>
      <AnalyticsAuditLogStats
        tenantKey={tenantKey}
        mentorId={mentorId}
        userId={userId}
      />
    </div>
  );
}

The component renders the filters, table, and pagination only. The "Audit" header / description shown in the screenshot is host-page chrome — copy the <h2> + block above to match.

Switching the audited agent at runtime

When the audit view is part of an agent picker (e.g. an admin browsing multiple agents), pass selectedMentorId to override the default mentorId without re-mounting:

<AnalyticsAuditLogStats
  tenantKey={tenantKey}
  mentorId={mentorId}
  userId={userId}
  selectedMentorId={pickerSelection ?? mentorId}
/>

selectedMentorId takes precedence when set; otherwise mentorId is used.

Step 3: Use MCP Tools for Customization

get_component_info("AnalyticsAuditLogStats")
get_hook_info("useAuditLog")
get_api_query_info("useGetAuditLogsQuery")

<AnalyticsAuditLogStats> Props

Import from @iblai/iblai-js/web-containers.

The actor filter, date range, and action filter are managed internally by useAuditLog — there is no prop to seed them.

What the tab renders

• Filters bar — actor combobox (populated from the current page's actor_email values), date-range picker, action select (All Actions / Create / Update / Delete).
• Table — USER, ACTION, TIME (relative, formatDistanceToNow).
• Action humanization — single-field boolean toggles render as Enabled <field> on <target> / Disabled <field> on <target>; multi-field updates list the changed fields with a built-in label map (e.g. enable_multi_query_rag → multi-query RAG, llm_name → LLM model). Unknown fields fall back to a snake_case → human-readable transform with common abbreviations (LLM, RAG, API, …) preserved in uppercase.
• Resource targets — mentorsettings is shown as mentor settings, trailing UUID is preserved so name collisions stay distinguishable (Settings for Foo (fbac00ea-...)).
• Pagination — 20 per page via IblPagination, server-side via limit/offset.
• Empty / error / 403 states — the component handles all three. 403 renders a card with "You do not have permission to view audit logs."

Custom UI: useAuditLog hook

Re-use the same query-state machinery the tab uses (server-side filters, action-code → name normalization, page/total math) when building a custom view:

import { useAuditLog } from "@iblai/iblai-js/web-containers";

const {
  data, isLoading, error,
  page, setPage, totalPages,
  actionFilter, setActionFilter,
  actorFilter, setActorFilter,
  dateRange, setDateRange,
} = useAuditLog({ tenantKey, userId, mentorId });

actionFilter is the integer code (0=create, 1=update, 2=delete) that the backend filter accepts, not the string the response returns. The hook normalizes both directions (the backend's create filter is currently dropped server-side, so the hook re-filters client-side as a safety net).

Related Exports

From @iblai/iblai-js/web-containers:

• AnalyticsAuditLogStats — the tab component.
• useAuditLog — the query/state hook.
• AuditLogDateRange — { from?: Date; to?: Date }.

From @iblai/data-layer:

• useGetAuditLogsQuery — the underlying RTK Query hook.
• AuditLogEntry, AuditLogsParams, AuditLogsResponse — payload types.

Step 4: Verify

Run /iblai-vibe-ops-test before telling the user the work is ready:

1. pnpm build -- must pass with zero errors
2. pnpm test -- vitest must pass
3. Touch test /agents/<id>/audit in the browser as /iblai-vibe-ops-test says: open it, use every control this skill added, read the console, check the phone width.

Important Notes

• Redux store: Must include mentorReducer and mentorMiddleware
• initializeDataLayer(): 5 args (v1.2+)
• @reduxjs/toolkit: Deduplicated via webpack aliases in next.config.ts
• Peer deps: sonner and @iblai/iblai-web-mentor must be installed (pnpm add sonner @iblai/iblai-web-mentor)
• No AgentSettingsProvider: Unlike AgentHistoryTab / AgentLLMTab / McpTab, this component takes raw props. If your app already mounts AgentSettingsProvider, you can read its values via useAgentSettings() in the page wrapper and forward them.
• RBAC: Ibl.Analytics/Core/read (viewer) or Ibl.* (organization admin). The 403 state already covers users without permission — do not gate the route at the layout level; let the component render the explanatory message.
• Brand guidelines: BRAND.md

Audit Logs REST API

For custom UI beyond <AnalyticsAuditLogStats>. All endpoints are prefixed with ${dmUrl}/api/ai-mentor/orgs/{org}/users/{user_id}/ where dmUrl is NEXT_PUBLIC_API_BASE_URL. Auth: Authorization: Token <token>.

Query parameters

Response shape

{
  "count": 137,
  "next": "...",
  "previous": null,
  "results": [
    {
      "id": 42,
      "action": "update",
      "actor_email": "conrad@ibleducation.com",
      "resource_type": "mentorsettings",
      "resource_id": 9,
      "resource_repr": "Settings for E2E Mentor 123 (b7af46ed-...)",
      "changes": {
        "enable_multi_query_rag": ["False", "True"],
        "llm_name": ["gpt-4o-mini", "gpt-4o"]
      },
      "timestamp": "2026-05-05T22:14:09Z"
    }
  ]
}

• action in the response is the string form (create / update / delete); the ?action= filter takes the integer code.
• changes is { [field]: [oldValue, newValue] }. Booleans serialize as the Python strings "True" / "False".
• resource_repr ends with (uuid) for agent-scoped resources — preserve it so name collisions remain disambiguated.

Known quirks

• ?action=0 (Create) is currently dropped server-side. The useAuditLog hook re-filters the response client-side to compensate; any custom UI should do the same until the backend fix lands.
• The trailing UUID in resource_repr is the agent's unique_id, not its primary key — useful for cross-referencing with agent settings APIs that key on unique_id.

Common errors

• 403 Forbidden — caller lacks Ibl.Analytics/Core/read. The tab component handles this state automatically; raw consumers should detect error.status === 403 and surface a permission notice rather than treating it as a generic failure.