hmem-mcp 7.2.2 → 7.2.4

package/package.json

@@ -1,7 +1,7 @@
 {
   "name": "hmem-mcp",
-  "version": "7.2.2",
-  "description": "Humanlike memory for AI agents — MCP server with 5-level lazy-loaded SQLite memory",
+  "version": "7.2.4",
+  "description": "Persistent memory and agent lifecycle for Claude Code — because sessions shouldn't start from zero.",
   "author": "Bumblebiber",
   "license": "MIT",
   "repository": {

package/skills/hmem-activate/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: hmem-activate
-description: Activate a project mid-session via load_project, then check for and fix misrouted O-entry exchanges. Use whenever the user says "aktiviere Projekt X", "lade Projekt X", "load project X", "wechsel zu Projekt X", or switches project context during a session.
+description: Activate a project mid-session via load_project, then check for and fix misrouted O-entry exchanges. Use whenever the user says "aktiviere Projekt X", "lade Projekt X", "load project X", "wechsel zu Projekt X", or switches project context during a session. ALSO use when you are about to call load_project mid-session for any reason — even if the user didn't explicitly ask to switch projects.
 ---
 
 # hmem-activate
@@ -13,8 +13,8 @@ Use when the user switches to a different project mid-session.
 Before calling load_project, check which project is currently active. The active project is shown in every tool response as `Active project: PXXXX <name>`.
 
 Note:
-- **OLD_PROJECT**: currently active project ID (e.g. P0043)
-- **OLD_O**: corresponding O-entry (same seq: P0043 → O0043)
+- **OLD_PROJECT**: currently active project ID (e.g. P0043) — or `none` if no project was active
+- **OLD_O**: corresponding O-entry (same seq: P0043 → O0043) — if OLD_PROJECT = `none`, use **O0000** (catch-all for unattributed sessions)
 - **TARGET_PROJECT**: the project the user wants to switch to
 - **TARGET_O**: corresponding O-entry (e.g. P0048 → O0048)
 

package/skills/hmem-release/SKILL.md

@@ -138,6 +138,7 @@ git push
 | cli.ts (new commands) | **hmem-update** (new CLI commands for users), hmem-setup |
 | cli-checkpoint.ts | hmem-config (checkpoint docs), hmem-read (summary docs) |
 | cli-log-exchange.ts | hmem-setup (hook docs) |
-| cli-context-inject.ts | hmem-wipe, hmem-setup |
+| cli-context-inject.ts | hmem-wipe, **hmem-setup**, **hmem-update** |
+| cli-hook-startup.ts | **hmem-setup** (first-message context docs), **hmem-update** (user-visible behavior) |
 | Any new skill added | **hmem-update** (list of skills to sync) |
 | Any new user-visible behavior | **hmem-update** — always |

package/skills/hmem-setup/SKILL.md

@@ -78,10 +78,10 @@ Script: `~/.claude/hooks/hmem-log-exchange.sh`
 Fires only after `/clear` (matcher: `"clear"`). Two hooks run in sequence:
 
 **a) `hmem context-inject`** — outputs `additionalContext` containing:
-  - Last 20 user/assistant messages from the pre-clear transcript
-  - Active project briefing (title + overview)
-  - Recent O-entries (session logs) linked to the project
-  - R-entries (rules)
+  - 5 most recently edited P-entries (one line each, active marked with `[*]`)
+  - Hint to call `read_memory({prefix:"P", titles_only:true})` for the full list
+  - All R-entries (rules, one line each)
+  - Footer with `load_project` call for the active project (if any)
   - Keeps the agent oriented after a context reset without a full `read_memory()` call.
 
 **b) `hmem deactivate`** — resets the active project for the new session:

package/skills/hmem-update/SKILL.md

@@ -370,6 +370,36 @@ Agents update L4 after every code change (enforced via `hmem-subagent` POST-TASK
 
 ---
 
+## Step 2k: v7.2.2 — Richer Session-Start Context + I-Entry Active Device Block
+
+**Only needed when upgrading from < v7.2.2**
+
+### context-inject: 5 most recent projects instead of full list
+
+`hmem context-inject` (SessionStart[clear] hook) now shows only the **5 most recently edited P-entries** instead of all projects. A hint is appended:
+```
+(full list: read_memory({prefix:"P", titles_only:true}))
+```
+No config change needed — automatic.
+
+### hook-startup: H-entry title fix
+
+H-entries without a `title` field (entries whose content lives entirely in `level_1`) now display correctly. Previously they appeared as blank lines in the `--- Human context (H-entries) ---` block.
+
+### hook-startup: Active device block
+
+The first-message context injection now includes a **`--- Active device (I00XX) ---`** block showing:
+- I-entry body (L1) — device name, specs, IP
+- All L2 section titles (Specs, OS, IP Address, Access, Services, Apps, Storage, Notes, Rules)
+- For the **Apps** section: full L3 title list (installed software)
+- Irrelevant L2/L3 nodes are filtered out
+
+### hook-startup: recent projects limit raised 3 → 5
+
+The `--- Recent projects ---` block in the first-message injection now shows 5 most recently updated P-entries (was 3).
+
+---
+
 ## Step 3: Entry Migration
 
 Some versions introduce new data formats. Check if migration is needed: