merge: AgentProject membership — the junction is now load-bearing

Merges feat/agent-project-membership (task 166/152). An agent can finally join
many projects: membership routes, per-membership role and optional brain
override, chat resolving context from the membership, and CLI attach-agent /
detach-agent / get memberships. The legacy Agent.projectId fallback is kept for
this release as the schema comment promises.

Conflict resolved in main.ts: that branch added its RBAC case to
mapUrlToPermission inline, while this branch had extracted the function to
middleware/url-permission.ts so it could be tested. Kept the extraction and
ported the case into it — joining a team is `edit:agents:<name>`, not
`create:agents`, which is the authority to mint a whole agent.

Also fixes something the extraction makes possible. security.test.ts contained a
hand-copied reproduction of mapUrlToPermission, because main.ts self-invokes
main() and nothing in it was importable. A security test asserting against a
copy stops testing the system the moment the real one changes, and this copy had
already drifted. It now imports the real function; all 18 assertions still pass,
against the code that actually runs.

Local: build clean, workspace 2550 passed, lint unchanged at 794 errors.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01N4wNHWf7xSwnZCWpJcyv9p

.taskmaster/tasks/task_032.md

@@ -0,0 +1,70 @@
+# Task ID: 32
+
+**Title:** Implement smart context optimization
+
+**Status:** done
+
+**Dependencies:** 29 ✓
+
+**Priority:** medium
+
+**Description:** Add token counting and decision logic to intelligently skip LLM filtering when responses are small enough, and cache filtering decisions for repeated queries.
+
+**Details:**
+
+1. Create `src/mcplocal/src/llm/token-counter.ts`:
+   ```typescript
+   export function estimateTokens(text: string): number {
+     // Simple estimation: ~4 chars per token for English
+     // More accurate: use tiktoken or similar library
+     return Math.ceil(text.length / 4);
+   }
+   ```
+2. Create `src/mcplocal/src/llm/filter-cache.ts`:
+   ```typescript
+   export class FilterCache {
+     private cache: LRUCache<string, FilterDecision>;
+     
+     shouldFilter(toolName: string, params: unknown, responseSize: number): boolean {
+       const key = this.computeKey(toolName, params);
+       const cached = this.cache.get(key);
+       if (cached) return cached.shouldFilter;
+       // No cache hit - use default threshold logic
+       return responseSize > this.tokenThreshold;
+     }
+     
+     recordDecision(toolName: string, params: unknown, decision: FilterDecision): void {
+       const key = this.computeKey(toolName, params);
+       this.cache.set(key, decision);
+     }
+   }
+   ```
+3. Add configuration options:
+   - `tokenThreshold`: number (default 1000 tokens)
+   - `filterCacheSize`: number (default 1000 entries)
+   - `filterCacheTtl`: number (default 3600 seconds)
+4. Integrate into LlmProcessor:
+   ```typescript
+   async filterResponse(...) {
+     const tokens = estimateTokens(JSON.stringify(rawResponse));
+     if (tokens < this.config.tokenThreshold) {
+       // Not worth filtering - return as-is
+       return { filtered: false, response: rawResponse };
+     }
+     // Proceed with LLM filtering
+   }
+   ```
+5. Add metrics tracking:
+   - Total tokens processed
+   - Tokens saved by filtering
+   - Filter cache hit rate
+   - Average latency added by filtering
+
+**Test Strategy:**
+
+1. Unit test: Token estimation is reasonably accurate
+2. Unit test: Cache correctly stores and retrieves decisions
+3. Unit test: Threshold logic skips filtering for small responses
+4. Unit test: Cache TTL expiration works correctly
+5. Integration test: Metrics are recorded accurately
+6. Performance test: Cache improves latency for repeated queries