Update docs for agent rename and KV changes

Author: chriswritescode-dev

packages/memory/CHANGELOG.md

@@ -10,7 +10,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Added
 
 - Project KV store for ephemeral project state with TTL management
-- `memory-kv-set`, `memory-kv-get`, `memory-kv-delete`, `memory-kv-list` tools for managing project state
+- `memory-kv-set`, `memory-kv-get`, `memory-kv-list` tools for managing project state
 - Automatic cleanup of expired KV entries (30-minute interval)
 - Default 24-hour TTL for KV entries
 
@@ -23,9 +23,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Added
 
 - `experimental.chat.messages.transform` hook: automatically injects project conventions and decisions into system prompts for every LLM call, with configurable token budget and caching
-- Skill tool awareness in Code and Architect agent prompts
-- `plan_enter` permission on Code agent for switching to Architect mode
-- `memory-edit` documentation in Memory agent's tool list
+- Skill tool awareness in code and architect agent prompts
+- `plan_enter` permission on code agent for switching to architect mode
+- `memory-edit` documentation in Librarian agent's tool list
 - Agent name logging in all tool handlers via new ToolContext.agent field
 
 ### Changed
@@ -35,24 +35,24 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
-- Memory agent tool documentation now lists all 7 available tools (was missing memory-edit)
+- Librarian agent tool documentation now lists all 7 available tools (was missing memory-edit)
 
 ## [0.0.9] - 2026-02-27
 
 ### Added
 
-- `experimental.chat.messages.transform` hook that injects read-only enforcement reminder into Architect agent sessions, preventing file edits and non-readonly tool usage at the message level
-- Code Review agent (`ocm-code-review`) — read-only subagent for convention-aware code reviews with memory integration, invoked via Task tool
-- `/review` command that triggers the Code Review agent to review current changes
+- `experimental.chat.messages.transform` hook that injects read-only enforcement reminder into architect agent sessions, preventing file edits and non-readonly tool usage at the message level
+- auditor agent (`ocm-auditor`) — read-only subagent for convention-aware code reviews with memory integration, invoked via Task tool
+- `/review` command that triggers the auditor agent to review current changes
 
 ### Changed
 
-- Restricted `memory-planning-update` and `memory-planning-search` tools to Memory subagent only — Code and Architect agents now delegate planning operations via @Memory Task tool
-- Overhauled Code and Architect agent system prompts with tone/style guidelines, tool usage policies, task management instructions, and planning state delegation patterns
+- Restricted `memory-planning-update` and `memory-planning-search` tools to librarian subagent only — code and architect agents now delegate planning operations via @librarian Task tool
+- Overhauled code and architect agent system prompts with tone/style guidelines, tool usage policies, task management instructions, and planning state delegation patterns
 - `memory-plan-execute` now accepts optional `objective`, `phases`, and `findings` parameters and saves planning state inline before dispatching the plan, eliminating the need for a separate `memory-planning-update` call
-- Planning instruction appended to dispatched plans now directs Code agent to delegate planning updates to @Memory subagent
-- Updated Memory agent description to include planning state and session progress management
-- Updated Code Review agent description to accurately reflect its capabilities
+- Planning instruction appended to dispatched plans now directs code agent to delegate planning updates to @librarian subagent
+- Updated librarian agent description to include planning state and session progress management
+- Updated auditor agent description to accurately reflect its capabilities
 
 ## [0.0.6] - 2026-02-24
 
@@ -62,9 +62,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Planning state tools: `memory-planning-update` and `memory-planning-get` for tracking session objectives, phases, findings, and errors
 - `memory-plan-execute` tool for creating new Code sessions with approved implementation plans
 - Three embedding providers: local (`all-MiniLM-L6-v2`), OpenAI (`text-embedding-3-small/large`, `ada-002`), and Voyage (`voyage-code-3`, `voyage-2`)
-- Bundled Code agent (`ocm-code`) with memory-aware coding workflows
-- Bundled Architect agent (`ocm-architect`) for read-only planning with automatic plan handoff
-- Bundled Memory agent (`ocm-memory`) for expert knowledge curation and post-compaction extraction
+- Bundled code agent (`ocm-code`) with memory-aware coding workflows
+- Bundled architect agent (`ocm-architect`) for read-only planning with automatic plan handoff
+- Bundled librarian agent (`ocm-librarian`) for expert knowledge curation and post-compaction extraction
 - Compaction context injection with custom prompt, planning state, conventions, and decisions
 - Configurable compaction settings: custom prompt, inline planning, token budget, snapshot storage
 - CLI export/import for backing up and migrating memories as JSON or Markdown

packages/memory/README.md

@@ -34,7 +34,7 @@ The local embedding model downloads automatically on install. For API-based embe
 - **Compaction Context Injection** - Injects conventions and decisions into session compaction for seamless continuity
 - **Automatic Memory Injection** - Injects relevant project memories into user messages via semantic search with distance filtering and caching
 - **Project KV Store** - Ephemeral key-value storage with TTL management for project state
-- **Bundled Agents** - Ships with Code, Architect, and Memory agents preconfigured for memory-aware workflows
+- **Bundled Agents** - Ships with Code, Architect, and Librarian agents preconfigured for memory-aware workflows
 - **CLI Tools** - Export, import, list, stats, and cleanup commands via `ocm-mem` binary
 - **Dimension Mismatch Detection** - Detects embedding model changes and guides recovery via reindex
 
@@ -44,14 +44,14 @@ The plugin bundles four agents that integrate with the memory system:
 
 | Agent | ID | Mode | Description |
 |-------|----|------|-------------|
-| **Code** | `ocm-code` | primary | Primary coding agent with memory awareness. Checks memory before unfamiliar code, stores architectural decisions and conventions as it works. Delegates planning operations to @Memory subagent. |
-| **Architect** | `ocm-architect` | primary | Read-only planning agent. Researches the codebase, delegates to @Memory for broad knowledge retrieval, designs implementation plans, then hands off to Code via `memory-plan-execute`. |
-| **Memory** | `ocm-memory` | subagent | Expert agent for managing project memory. Handles post-compaction memory extraction and contradiction resolution. |
-| **Code Review** | `ocm-code-review` | subagent | Read-only code reviewer with access to project memory for convention-aware reviews. Invoked via Task tool to review diffs, commits, branches, or PRs against stored conventions and decisions. |
+| **code** | `ocm-code` | primary | Primary coding agent with memory awareness. Checks memory before unfamiliar code, stores architectural decisions and conventions as it works. Delegates planning operations to @librarian subagent. |
+| **architect** | `ocm-architect` | primary | Read-only planning agent. Researches the codebase, delegates to @librarian for broad knowledge retrieval, designs implementation plans, then hands off to code via `memory-plan-execute`. |
+| **librarian** | `ocm-librarian` | subagent | Expert agent for managing project memory. Handles post-compaction memory extraction and contradiction resolution. |
+| **auditor** | `ocm-auditor` | subagent | Read-only code auditor with access to project memory for convention-aware reviews. Invoked via Task tool to review diffs, commits, branches, or PRs against stored conventions and decisions. |
 
-The Code Review agent is a read-only subagent (`temperature: 0.0`) that can read memory but cannot write, edit, or delete memories or execute plans. It is invoked by other agents via the Task tool to review code changes against stored project conventions and decisions.
+The auditor agent is a read-only subagent (`temperature: 0.0`) that can read memory but cannot write, edit, or delete memories or execute plans. It is invoked by other agents via the Task tool to review code changes against stored project conventions and decisions.
 
-The Architect agent operates in read-only mode (`temperature: 0.0`, all edits denied) with additional message-level read-only enforcement via the `experimental.chat.messages.transform` hook. After the user approves a plan, it calls `memory-plan-execute` which creates a new Code session with the full plan as context.
+The architect agent operates in read-only mode (`temperature: 0.0`, all edits denied) with additional message-level read-only enforcement via the `experimental.chat.messages.transform` hook. After the user approves a plan, it calls `memory-plan-execute` which creates a new code session with the full plan as context.
 
 ## Tools
 
@@ -74,7 +74,6 @@ Ephemeral key-value storage for project state with automatic TTL-based expiratio
 |------|-------------|
 | `memory-kv-set` | Store a value with optional TTL (default 24 hours) |
 | `memory-kv-get` | Retrieve a value by key |
-| `memory-kv-delete` | Delete a value by key |
 | `memory-kv-list` | List all active KV entries for the project |
 
 ## CLI
@@ -260,11 +259,11 @@ When enabled, logs are written to the specified file with timestamps. The log fi
 #### Execution
 - `executionModel` - Model override for plan execution sessions, format: `provider/model` (e.g. `anthropic/claude-haiku-3-5-20241022`). When set, `memory-plan-execute` uses this model for the new Code session. When empty or omitted, OpenCode's default model is used (typically the `model` field from `opencode.json`). **Recommended:** Set this to a fast, cheap model (e.g. Haiku or MiniMax) and use a smart model (e.g. Opus) for the Architect session — planning needs reasoning, execution needs speed.
 
-## Architect → Code Workflow
+## architect → code Workflow
 
-Plan with a smart model, execute with a fast model. The Architect agent researches and designs; the Code agent implements.
+Plan with a smart model, execute with a fast model. The architect agent researches and designs; the code agent implements.
 
-Set `executionModel` in your config to a fast model (e.g., Haiku) and use a smart model (e.g., Opus) for the Architect session.
+Set `executionModel` in your config to a fast model (e.g., Haiku) and use a smart model (e.g., Opus) for the architect session.
 
 See the [full workflow guide](https://chriswritescode-dev.github.io/opencode-manager/features/memory/#architect--code) for setup details.