context-mode 1.0.65 → 1.0.66

package/skills/context-mode/SKILL.md

@@ -100,6 +100,7 @@ About to run a command / read a file / call an API?
 | Playwright console/network | `browser_*(filename)` → `ctx_execute_file(path)` | Save to file, analyze in sandbox |
 | MCP output (already in context) | Use directly | Don't re-index — it's already loaded |
 | MCP output (need multi-query) | `ctx_execute` to save → `ctx_index(path)` → `ctx_search` | Save to file first, index server-side |
+| Wipe indexed KB content | `ctx_purge(confirm: true)` | Permanently deletes all indexed content |
 
 ## Automatic Triggers
 
@@ -288,6 +289,7 @@ Subagents automatically receive context-mode tool routing via a PreToolUse hook.
 - Passing ANY large data to `ctx_index(content: ...)` → data enters context as a parameter. **Always** use `ctx_index(path: ...)` to read server-side. The `content` parameter should only be used for small inline text you're composing yourself.
 - Calling an MCP tool (Context7 `query-docs`, GitHub API, etc.) then passing the response to `ctx_index(content: response)` → **doubles** context usage. The response is already in context — use it directly or save to file first.
 - Ignoring `browser_navigate` auto-snapshot → navigation response includes a full page snapshot. Don't rely on it for inspection — call `browser_snapshot(filename)` separately.
+- Expecting `ctx_stats` to reset or wipe anything → `ctx_stats` is read-only (shows stats only). Use `ctx_purge(confirm: true)` to permanently delete all indexed content.
 
 ## Reference Files
 

package/skills/ctx-purge/SKILL.md

@@ -0,0 +1,35 @@
+---
+name: ctx-purge
+description: |
+  Purge the context-mode knowledge base. Permanently deletes all indexed content
+  and resets session stats. This is destructive and cannot be undone.
+  Trigger: /context-mode:ctx-purge
+user-invocable: true
+---
+
+# Context Mode Purge
+
+Permanently deletes ALL session data for this project: knowledge base, session events, analytics, and stats.
+
+## Instructions
+
+1. **Warn the user**: This is irreversible. Everything will be deleted:
+   - FTS5 knowledge base (all indexed content from `ctx_index`, `ctx_fetch_and_index`, `ctx_batch_execute`)
+   - Session events DB (analytics, metadata, resume snapshots)
+   - Session events markdown file
+   - In-memory session stats
+2. Call the `mcp__context-mode__ctx_purge` MCP tool with `confirm: true`.
+3. Report the result to the user — the response lists exactly what was deleted.
+
+## When to Use
+
+- When the KB contains stale or incorrect content polluting search results.
+- When switching between unrelated projects in the same session.
+- When you want a completely fresh start for this project.
+
+## Important
+
+- `ctx_purge` is the **only** way to delete session data. No other mechanism exists.
+- `ctx_stats` is read-only — shows statistics only.
+- `/clear` and `/compact` do NOT affect any context-mode data.
+- There is no undo. Re-index content if you need it again.

package/skills/ctx-stats/SKILL.md

@@ -3,6 +3,8 @@ name: ctx-stats
 description: |
   Show how much context window context-mode saved this session.
   Displays token consumption, context savings ratio, and per-tool breakdown.
+  Read-only — shows stats only, no reset capability.
+  To wipe the knowledge base entirely, use ctx_purge instead.
   Trigger: /context-mode:ctx-stats
 user-invocable: true
 ---
@@ -18,3 +20,7 @@ Show context savings for the current session.
 3. After the full output, add ONE sentence highlighting the key savings metric, e.g.:
    - "context-mode saved **12.4x** — 92% of data stayed in sandbox."
    - If no data yet: "No context-mode calls yet this session."
+
+## Purge
+
+- **`ctx_purge(confirm: true)`** — Permanently deletes all indexed content from the knowledge base. Use `/context-mode:ctx-purge` for this.

package/start.mjs

@@ -13,6 +13,13 @@ if (!process.env.CLAUDE_PROJECT_DIR) {
   process.env.CLAUDE_PROJECT_DIR = originalCwd;
 }
 
+// Platform-agnostic project dir — guaranteed to be set for ALL platforms.
+// Adapters may set their own env var (GEMINI_PROJECT_DIR, etc.) but this
+// is the universal fallback so server.ts getProjectDir() never relies on cwd().
+if (!process.env.CONTEXT_MODE_PROJECT_DIR) {
+  process.env.CONTEXT_MODE_PROJECT_DIR = originalCwd;
+}
+
 // Routing instructions file auto-write DISABLED for all platforms (#158, #164).
 // Env vars like CLAUDE_SESSION_ID may not be set at MCP startup time, making
 // the hook-capability guard unreliable. Writing to project dirs dirties git trees