[OpenSpec] hermiq-ai-tooling

[github-actions]

⚠️ OpenSpec-managed issue — this content is automatically synced
from the openspec/ directory. Manual edits will be overwritten on next sync.

Artifacts

• ✅ Metadata
• ✅ Proposal
• ✅ Specs
• ✅ Design
• ✅ Tasks
• 🔲 Plan
• 🔲 Review

Summary

Extends portaliq-mcp-adoption — it does not replace or repeat it. That change (see
openspec/changes/portaliq-mcp-adoption/) deletes the app-template scaffold and gives portaliq a
two-tool read surface on the portal inbox
(specs/portaliq-mcp-surface/spec.md#requirement-a-curated-read-only-dialect-on-the-portal-inbox).
This change takes the next step the product framing asks for: every internal staff action
portaliq offers becomes an MCP tool, so any of them can in principle be automated by an AI agent
— and every one of them is governed by hermiq's per-agent grant model (scope
read/create/update/delete × reach self/user/instance/external, default-deny
writes, human approval gates, audit trail). Even without automation, this is what makes chat a way
of commanding the app: an operator chats while portaliq executes.

Two hard lines are stated normatively, not as intentions: writes are curated tools only (the
schema-derived catalog stays read-only, preserving portaliq-mcp-adoption's
#requirement-a-curated-read-only-dialect-on-the-portal-inbox reasoning), and MCP tools serve
internal staff and their agents only — the surface is never reachable from, or exposed at, the
external portal edge (ADR-046).

Specs

• 📄 portaliq-mcp-surface

Tasks

• Implement
• Test
• Implement
• Test
• Implement
• Test
• Implement
• Test
• Implement
• Test
• Implement
• Test

Design

See design.md for technical design details.

---

Synced from openspec/changes/hermiq-ai-tooling by OpenSpec workflow
App: portaliq