docs: add MCP Interceptors research note

docs/superpowers/plans/2026-09-18-mcp-interceptors-research-note.md

@@ -0,0 +1,27 @@
+# MCP Interceptors Research Note Implementation Plan
+
+> **For agentic workers:** Execute this plan inline with validation checkpoints.
+
+**Goal:** Add and publish a sourced research note on MCP Interceptors as reusable policy and context mediation.
+
+**Architecture:** Explain validator/mutator primitives, lifecycle hooks, trust-boundary execution, chain ordering, audit mode, and in-process/sidecar/remote deployment. Map them to CF gateways, proxies, authorization, redaction, validation, and audit.
+
+**Tech Stack:** Markdown, YAML frontmatter, Devbox, Git, GitHub CLI.
+
+---
+
+### Task 1: Write `research/mcp-interceptors.md`
+
+- [ ] Add frontmatter with title `MCP Interceptors: Reusable Policy and Context Mediation`, author `Ruben Koster (@rkoster)`, date `2026-09-18`, tags `[authorization, observability-governance, inter-agent-comms, ecosystem-survey]`, `cf_areas: [uaa, capi, diego, loggregator]`, `status: draft`, ratings, and charter/repository/SEP sources.
+- [ ] Explain the M x N problem caused by bespoke sidecars, proxies, and gateways for cross-cutting agent concerns.
+- [ ] Cover validator and mutator types, lifecycle hooks for tool calls, resource reads, prompts, sampling, elicitation, and extension to LLM/custom workflows.
+- [ ] Cover trust-boundary-aware execution, priority-ordered chains, audit mode, and in-process/sidecar/remote deployment trade-offs.
+- [ ] Identify sample use cases such as PII redaction, schema validation, and audit logging, and distinguish the experimental proposal from a finalized MCP standard.
+- [ ] Assess CF relevance for gateways, sidecars, service proxies, UAA/policy, Loggregator, multi-tenant ordering, and platform/application ownership.
+- [ ] Add open questions about ordering, failure behavior, mutation accountability, trust, latency, sensitive data, and interceptor discovery.
+
+### Task 2: Validate and publish
+
+- [ ] Run `devbox run validate`, `devbox run test`, and `git diff --check`.
+- [ ] Stage only the note and approved spec/plan, commit `docs: add MCP Interceptors research note`, push `research/mcp-interceptors`, and open a checklist-complete PR targeting `main`.
+- [ ] Verify PR metadata and CI with `gh pr view`.

docs/superpowers/specs/2026-09-18-mcp-interceptors-research-note-design.md

@@ -0,0 +1,29 @@
+# MCP Interceptors Research Note Design
+
+## Goal
+
+Add a sourced research note on the MCP Interceptors Working Group's proposal for reusable
+policy and context mediation across agentic operations.
+
+## Scope
+
+The note will cover validator and mutator interceptor types, lifecycle hooks for tools,
+resources, prompts, sampling, and elicitation, extensibility to LLM/custom workflows,
+trust-boundary-aware execution, priority chains, audit mode, and in-process/sidecar/remote
+deployment. It will identify the working group's draft/experimental maturity.
+
+The Cloud Foundry analysis will map interceptors to gateways, sidecars, authorization, PII
+redaction, schema validation, audit logging, service proxies, and Loggregator. It will not claim
+existing CF/MCP Interceptors integration.
+
+## Structure and evidence
+
+Create `research/mcp-interceptors.md` with the required four sections and frontmatter. Use the
+official charter, experimental extension repository, SEP references, and MCP lifecycle context.
+Clearly distinguish protocol-level interceptor interfaces from host-specific hooks and general
+middleware.
+
+## Validation
+
+Run Devbox validation and tests, inspect whitespace/staged files, commit the note and plan on
+`research/mcp-interceptors`, push, and open a PR targeting `main` without unrelated artifacts.

presentation/.gitignore

@@ -0,0 +1,2 @@
+node_modules/
+build/

presentation/README.md

@@ -0,0 +1,46 @@
+# Agentic Runtime Summit Deck
+
+Ten-minute Cloud Foundry Summit working-group update: **from research to
+sandboxing POCs**.
+
+## Run locally
+
+```bash
+cd presentation
+npm install
+npm run slides
+```
+
+`npm run slides` rebuilds on every save and opens a live-reloading preview
+window. If you have no GUI browser available, use `npm run slides:watch`
+instead — it watches and rebuilds `build/preview.html` without opening anything.
+
+Export deliverables with:
+
+```bash
+cd presentation
+npm run slides:pdf
+npm run slides:html
+```
+
+PDF export needs a Chromium-compatible browser.
+
+- **Linux:** Chromium is provided by Devbox. Nothing to install.
+- **macOS:** Devbox excludes Chromium on Darwin, because the Nix package does
+  not evaluate on Apple Silicon. Marp uses a locally installed Chrome, Edge, or
+  Firefox instead — install one if you don't already have it, or point
+  `CHROME_PATH` at an existing browser binary.
+
+Generated files are written to `build/`, which is intentionally local output.
+
+## Talk shape
+
+The deck starts with repository and working-group momentum, explains the agent
+loop at a conceptual level, and makes the local-tool execution boundary the
+reason sandboxing is the first POC focus. The demo slide is deliberately a
+placeholder until the POC leads select a scenario.
+
+The activity figures are a repository snapshot from 20 September 2026:
+63 PRs total, 43 merged, 20 open, and 3 open issues. OpenSandbox is described
+as researched prior art linked to an open AAIF project proposal, not as an
+adopted architecture.