SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsiblai-vibe-agent-privacy

iblai-vibe-agent-privacy

Agents & models

View repository

Updated 10-15 22:16

About this skill

/iblai-vibe-agent-privacy

Add the agent Privacy tab -- "Control what personal information the agent sees and how Incognito works for its conversations." Two sub-tabs behind one pill switcher:

  • PII Filtering — an inline PII filtering toggle; when on, what to do when PII is detected (Allow, Redact, Mask, Block), the Block Message, the Entity Types chips (Person, Email, Phone, SSN, Credit Card, Location, Date / Time, Passport, Driver's License, IP Address, IBAN, Medical License, Bank Number) and Also filter AI responses.
  • Incognito — one choice over how Incognito works for this agent: Users Decide, Always Incognito, or Never Incognito. This replaces the old Private Mode switch that lived on the Settings tab.

This is one tab in the wider agent-settings family. All tabs share the same AgentSettingsProvider wrapper (optional for this tab — identity can be passed as props).

Privacy — PII Filtering sub-tab

Privacy — Incognito sub-tab

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

Related Incognito surfaces

Incognito is one feature with four tiers; this tab is the agent tier. /iblai-vibe-incognito explains how they resolve and mounts the in-chat toggle.

  • /iblai-vibe-incognito — the chat-header Incognito toggle and its "Enable Incognito for this chat?" dialog, plus the tier order.
  • /iblai-vibe-profile — the user's own Privacy tab: their default (Normal / Anonymized / Incognito).
  • /iblai-vibe-account — organization settings → Advanced → "Allow users to control chat privacy", the gate that makes the Incognito sub-tab (and the toggle) exist at all.

Prerequisites

  • Auth must be set up first (/iblai-vibe-auth)
  • MCP server + skills configured (@iblai/mcp in .mcp.json)
  • AgentSettingsProvider wrapping the route (see /iblai-vibe-agent-setting Step 2), or pass tenantKey / mentorId / username as props.
  • Ask the user for a real mentorId (agent UUID). Do NOT invent one.
  • The Incognito sub-tab appears only when the organization allows user chat-privacy control (allow_user_chat_privacy_control). With the gate off there is no Incognito to govern, and the tab shows PII Filtering alone.

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 AgentPrivacyTab

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

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

export default function AgentPrivacyPage() {
  return (
    <div className="flex h-full flex-col bg-white">
      <AgentPrivacyTab />
    </div>
  );
}

The tab reads tenantKey, mentorId, and username from the nearest <AgentSettingsProvider>, loads the agent settings and the organization's chat-privacy gate, and decides which sub-tabs to offer (see Which sub-tabs appear below).

With identity overrides

When the tab is not rendered inside an <AgentSettingsProvider>, pass identity explicitly:

<AgentPrivacyTab
  tenantKey="my-tenant"
  mentorId="00000000-0000-0000-0000-000000000000"
  username="learner@example.com"
/>;

With RBAC and gated mutations

<AgentPrivacyTab
  enableRBAC
  executeGatedAction={(fn) => withConfirm(fn)}
/>;

enableRBAC honors field-level permissions from agent settings — each field (enable_privacy_router, privacy_action, privacy_response, privacy_entities, enable_privacy_output_filter, disable_chathistory, disable_privacy_mode) can be independently read-only. executeGatedAction wraps every save so the host app can interpose a confirmation or upgrade gate.

Step 3: Customize Labels (Optional)

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

<AgentPrivacyTab
  labels={{
    header: { title: "Data privacy", description: "Filter PII; decide how Incognito works." },
    subTabs: { piiFiltering: "Personal data", incognito: "Incognito" },
    incognito: {
      options: { always: { title: "Always private", description: "Nothing is ever saved." } },
    },
  }}
/>;

Step 4: Use MCP Tools for Customization

get_component_info("AgentPrivacyTab")
get_component_info("AgentSettingsProvider")
get_component_info("ChatPrivacyToggle")

<AgentPrivacyTab> Props

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

Prop Type Required Description
labels DeepPartial<PrivacyTabLabels> No Override user-visible strings (header, capability card, field labels, tooltips, action descriptions, entity labels, sub-tab names, the three Incognito cards, toasts)
tenantKey string No Identity override. Defaults to the nearest <AgentSettingsProvider>
mentorId string No Agent UUID override. Defaults to the provider
username string No Username override. Defaults to the provider
enableRBAC boolean No Honor per-field permissions from agent settings. Defaults to the provider value or false
executeGatedAction (fn: () => unknown) => unknown No Wrap each save mutation (e.g. confirmation/upgrade gate). Defaults to the provider value

What each sub-tab renders

PII Filtering

  • PII filtering capability card — the on/off switch sits inline at the top of the sub-tab with its explanation. When it is off the controls below stay visible but inert, so an admin can preview what they will get before turning it on.

  • When PII is detected — a select with an inline description of the chosen action:

    Action Wire value What happens
    Allow allow Detected but left untouched — the message passes through unchanged. For monitoring or auditing without altering messages
    Redact (default) redact Replaced with its type — e.g. Email [EMAIL_ADDRESS]
    Mask mask Hidden behind asterisks — e.g. Email ***********
    Block block The message is rejected and your block message is shown

  • Block Message — a textarea shown only for Block; saves on blur.

  • Entity Types — the 13 chips, multi-select. With none selected the hint reads "Using defaults."

  • Also filter AI responses — scrubs PII from the agent's output as well (may add slight latency to streamed responses).

Incognito

"Incognito keeps a conversation out of chat history and memory. Choose how it works for this agent." Three radio cards:

Card Flags saved Effect in the chat header
Users Decide disable_chathistory: false, disable_privacy_mode: false Users can turn Incognito on for any conversation; otherwise chats are saved as usual
Always Incognito disable_chathistory: true Every conversation is incognito, for anyone. The Incognito pill shows locked on
Never Incognito disable_privacy_mode: true Incognito isn't available for this agent; the toggle disappears. Users whose profile default is Incognito are saved anonymously instead

One control writes both flags in the same PUT, so the contradictory "always and never" pair cannot be configured from here. A record that still carries both (older data) behaves as Always Incognito: the backend checks disable_chathistory first, answers mode: disabled, source: mentor, and writes no history — even though the tab displays it as Never Incognito. Choosing any card rewrites both flags and clears the mismatch.

Which sub-tabs appear

Condition Result
Organization gate allow_user_chat_privacy_control off Incognito sub-tab not offered
enableRBAC and no PII field is readable (permissions.field) PII Filtering not offered
enableRBAC and either Incognito flag is unreadable Incognito not offered
Nothing readable The tab shows a single notice: "You don't have permission to view this agent's privacy settings."

Members typically see Incognito read-only and no PII Filtering. Writes stay per control: a field the user may read but not write renders disabled.

Settings fields

Read with GET and written with PUT on ${dmUrl}/api/ai-mentor/orgs/{org}/users/{username}/mentors/{mentor_unique_id}/settings/ (a partial update — send only what changes):

Field Type Sub-tab
enable_privacy_router boolean PII Filtering — the capability switch
privacy_action "allow" | "redact" | "mask" | "block" PII Filtering
privacy_response string PII Filtering — the block message
privacy_entities string[] (PERSON, EMAIL_ADDRESS, …) — [] means defaults PII Filtering
enable_privacy_output_filter boolean PII Filtering
disable_chathistory boolean Incognito — Always
disable_privacy_mode boolean | null (null = off on older records) Incognito — Never

Related Exports

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

  • AgentPrivacyTab, AgentPrivacyTabProps — this tab.
  • AGENT_PRIVACY_TAB_LABELS, PrivacyTabLabels — the default label bundle and its type (now including subTabs and incognito).

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

  • ChatPrivacyToggle — the chat-header toggle that reacts to the Incognito policy (/iblai-vibe-incognito).

From @iblai/iblai-js/data-layer:

  • PRIVACY_ACTIONS (['allow','redact','mask','block']), PRIVACY_ENTITY_TYPES, DEFAULT_PRIVACY_ACTION, DEFAULT_PRIVACY_RESPONSE, and the PrivacyAction / PrivacyEntityType types.
  • isPrivacyBlockedError, PRIVACY_BLOCKED_ERROR_KIND — recognise the privacy_blocked error a Block action sends over the chat stream, to style it apart from a moderation block.
  • useGetMentorSettingsQuery, useEditMentorJsonMutation — the read and the JSON PUT the tab uses.
  • useGetTenantChatPrivacyConfigQuery — the organization gate the Incognito sub-tab depends on.
  • MentorChatPrivacyPublicSettings — disable_privacy_mode as members read it from the agent's public settings.

The policy ↔ flags mapping helpers (INCOGNITO_POLICIES, incognitoPolicyFrom, incognitoFlags) live in the tab's source but are not re-exported from the package entry — custom UI should mirror the table above.

SDK source: packages/web-containers/src/components/modals/edit-mentor-modal/tabs/privacy-tab.tsx (with privacy-tab/incognito-policy.ts and privacy-tab/labels.ts).

Step 5: 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. Start dev server and touch test: 
    pnpm dev &
npx playwright screenshot http://localhost:3000/agents/<id>/privacy /tmp/agent-privacy.png

Stable selectors: privacy-sub-tab-pii, privacy-sub-tab-incognito, privacy-capability-toggle, privacy-incognito-{users|always|never} (with aria-pressed), privacy-no-access.

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)
  • Provider optional: this tab uses useAgentSettingsOptional, so it works inside AgentSettingsProvider or with explicit identity props.
  • JSON settings endpoint: all fields are persisted via the JSON /settings/ mutation (useEditMentorJsonMutation), not the multipart variant, so privacy_entities round-trips cleanly. The tab refetches agent settings after each save because that mutation does not invalidate the settings cache tag.
  • Incognito saves refresh the chat header: choosing a card also invalidates ChatPrivacyEffective and the agent's mentorPublicSettings caches, so an open chat's Incognito pill locks, unlocks or disappears without a reload. Custom UI writing the two flags must do the same.
  • Private Mode moved: the switch that used to sit on the Settings tab's capabilities is gone; the Incognito sub-tab here is its replacement, and the user-facing name is now Incognito everywhere.
  • Brand guidelines: BRAND.md