aia 1.1.0 → 2.0.0.0.pre.alpha

data/docs/directives-reference.md

@@ -1,6 +1,121 @@
+<!-- Tocer[start]: Auto-generated, don't remove. -->
+
+## Table of Contents
+
+- [Directives Reference](#directives-reference)
+  - [Quick Reference](#quick-reference)
+  - [Directive Syntax](#directive-syntax)
+  - [Configuration Directives](#configuration-directives)
+    - [`/config`](#config)
+    - [`/model`](#model)
+    - [`/temperature`](#temperature)
+    - [`/top_p`](#top_p)
+    - [`/cost`](#cost)
+  - [File and Web Directives](#file-and-web-directives)
+    - [`/include`](#include)
+    - [`/paste`](#paste)
+    - [`/webpage`](#webpage)
+    - [`/skill`](#skill)
+    - [`/skills`](#skills)
+  - [Execution Directives](#execution-directives)
+    - [`/ruby`](#ruby)
+    - [`/say`](#say)
+    - [`/concurrent`](#concurrent)
+    - [`/verify`](#verify)
+    - [`/decompose`](#decompose)
+    - [`/debate`](#debate)
+    - [`/delegate`](#delegate)
+    - [`/spawn`](#spawn)
+    - [`/add_recruit`](#add_recruit)
+    - [`/reskill`](#reskill)
+    - [`/drop_recruit`](#drop_recruit)
+  - [Utility Directives](#utility-directives)
+    - [`/tools`](#tools)
+    - [`/next`](#next)
+    - [`/pipeline`](#pipeline)
+    - [`/mcp`](#mcp)
+    - [`/robots`](#robots)
+    - [`/terse`](#terse)
+    - [`/robot`](#robot)
+  - [Context Management Directives](#context-management-directives)
+    - [`/checkpoint`](#checkpoint)
+    - [`/restore`](#restore)
+    - [`/clear`](#clear)
+    - [`/review`](#review)
+    - [`/checkpoints_list`](#checkpoints_list)
+  - [TrakFlow Directives](#trakflow-directives)
+    - [`/tasks`](#tasks)
+    - [`/plan`](#plan)
+    - [`/task`](#task)
+  - [Model and Information Directives](#model-and-information-directives)
+    - [`/available_models`](#available_models)
+    - [`/compare`](#compare)
+    - [`/help`](#help)
+  - [Directive Processing Order](#directive-processing-order)
+  - [Advanced Usage Patterns](#advanced-usage-patterns)
+    - [Combining Directives](#combining-directives)
+    - [Dynamic Configuration](#dynamic-configuration)
+    - [Conditional Execution](#conditional-execution)
+    - [Workflow Automation](#workflow-automation)
+  - [Error Handling](#error-handling)
+    - [Common Errors](#common-errors)
+    - [Custom Directives](#custom-directives)
+    - [Best Practices](#best-practices)
+  - [Security Considerations](#security-considerations)
+    - [Safe Usage Tips](#safe-usage-tips)
+  - [Environment Variables for Directives](#environment-variables-for-directives)
+  - [Related Documentation](#related-documentation)
+
+<!-- Tocer[finish]: Auto-generated, don't remove. -->
+
 # Directives Reference
 
-Directives are special commands embedded in prompts that provide dynamic functionality. All directives begin with `/` and are processed before the prompt is sent to the AI model.
+Directives are slash commands available during interactive chat sessions. All directives begin with `/` and are processed before the prompt is sent to the AI model.
+
+Use `/help` at any time to see all available directives.
+
+## Quick Reference
+
+| Directive | Aliases | Description |
+|-----------|---------|-------------|
+| `/config` | `/cfg` | View or set configuration values |
+| `/model` | | View or change the AI model |
+| `/temperature` | `/temp` | Set temperature parameter |
+| `/top_p` | `/topp` | Set top_p parameter |
+| `/cost` | | Dump session cost/token metrics as CSV |
+| `//include` | `//import` | Include file content at prompt load time (prompt_manager, not chat-time) |
+| `/webpage` | `/website`, `/web` | Fetch and include web content |
+| `/skill` | | Include a Claude Code skill |
+| `/skills` | | List available Claude Code skills |
+| `/paste` | `/clipboard` | Paste from system clipboard |
+| `/ruby` | `/rb` | Execute Ruby code |
+| `/say` | | Text-to-speech |
+| `/concurrent` | `/conc` | Concurrent MCP server access |
+| `/verify` | | Verification network (two answers + reconciliation) |
+| `/decompose` | | Decompose prompt into parallel sub-tasks |
+| `/debate` | | Multi-round debate between robots |
+| `/delegate` | `/del` | Delegate subtasks via TrakFlow plan |
+| `/spawn` | | Spawn a one-shot specialist robot for the next prompt |
+| `/add_recruit` | `/add` | Recruit a persistent, `@mention`-able robot (optionally skilled) into the crew |
+| `/drop_recruit` | `/drop` | Remove a robot from the crew |
+| `/reskill` | | Reset a crew member to a clean slate and give it a new skill/role |
+| `/tools` | | List available tools |
+| `/mcp` | | MCP server connection status |
+| `/robots` | | Show active robot configuration |
+| `/robot` | | ASCII robot art |
+| `/help` | | Show directive help |
+| `/checkpoint` | `/ckp`, `/cp` | Create a conversation checkpoint |
+| `/restore` | | Restore to a checkpoint |
+| `/clear` | | Clear conversation context |
+| `/review` | `/context` | Review conversation with checkpoint markers |
+| `/checkpoints_list` | `/checkpoints` | List all checkpoints |
+| `/tasks` | `/tf` | Show TrakFlow ready tasks |
+| `/plan` | | Create a TrakFlow plan |
+| `/task` | | Create a TrakFlow task |
+| `/available_models` | `/am`, `/available`, `/all_models`, `/models`, `/llms` | List available AI models |
+| `/compare` | `/cmp` | Compare responses from multiple models |
+| `next` | | YAML front matter key: set next prompt in workflow (prompt_manager) |
+| `pipeline` | `workflow` | YAML front matter key: define prompt workflow sequence (prompt_manager) |
 
 ## Directive Syntax
 
@@ -10,10 +125,10 @@ Directives are special commands embedded in prompts that provide dynamic functio
 
 Examples:
 ```markdown
-/config model gpt-4
-/include my_file.md
-
-<%= "Hello World" %>
+/config temperature 0.8
+/model gpt-4o, claude-sonnet-4
+/debate
+Should we use a monolith or microservices?
 ```
 
 ## Configuration Directives
@@ -95,28 +210,51 @@ Set nucleus sampling parameter (alternative to temperature).
 
 **Aliases**: `/topp`
 
+### `/cost`
+Dump session cost and token metrics as CSV.
+
+**Syntax**: `/cost`
+
+**Example Output**:
+```
+model,input_tokens,output_tokens,total_tokens,cost,elapsed,similarity
+gpt-4o,1234,567,1801,$0.01234,2.3s,ref
+claude-sonnet-4-20250514,1234,890,2124,$0.00987,3.1s,87.2%
+TOTAL,2468,1457,3925,$0.02221,3.1s,
+```
+
+**Features**:
+- Shows per-turn breakdown: model, token counts, cost, elapsed time
+- Includes similarity scores when multi-model mode is active
+- Appends CSV to output file when `--output` is configured
+- Totals row at the bottom summarizes the session
+
 ## File and Web Directives
 
 ### `/include`
-Include content from files or websites.
+Include content from files into a prompt file.
 
-**Syntax**: `/include path_or_url`
+> **Note**: `/include` (also written as `//include` in prompt files) is handled by the
+> `prompt_manager` gem **during prompt loading**, before AIA processes the prompt.
+> It is NOT a chat-time directive — you cannot type `/include` at the chat prompt and
+> have it evaluated interactively. The inclusion happens when the prompt file is read
+> from disk, so the included content is already present by the time AIA sees the prompt.
 
-**Examples**:
+**Syntax** (inside a `.md` prompt file): `//include path_or_url`
+
+**Examples** (in prompt files):
 ```markdown
-/include README.md
-/include /path/to/config.yml
-/include ~/Documents/notes.md
-/include https://example.com/page
+//include README.md
+//include /path/to/config.yml
+//include ~/Documents/notes.md
 ```
 
 **Features**:
 - Supports tilde (`~`) and environment variable expansion
 - Prevents circular inclusions
-- Can include web pages (requires PUREMD_API_KEY)
 - Handles both absolute and relative file paths
 
-**Aliases**: `/import`
+**Aliases**: `//import`
 
 ### `/paste`
 Insert content from the system clipboard.
@@ -163,66 +301,78 @@ Include an AIA skill into the conversation context.
 **Examples**:
 ```markdown
 /skill code-quality          # Exact match
-/skill code                  # Prefix match (finds first match starting with "code")
+/skill code                  # Prefix match (finds first alphabetical match starting with "code")
 /skill frontend-design       # Exact match
 /skill front                 # Prefix match (finds "frontend-design")
 ```
 
 **Features**:
-- Reads `SKILL.md` from the skills directory (default: `~/.prompts/skills/<skill_name>/`)
+- Reads `SKILL.md` from `<skills_dir>/<skill_name>/`
+- Skills directory: `AIA.config.skills.dir` (default `~/.prompts/skills`); if `AIA_PROMPTS__SKILLS_PREFIX` is set, appended to `AIA_PROMPTS__DIR` to form the base
 - Supports prefix matching: `/skill code` finds the first subdirectory starting with "code"
 - Exact matches take priority over prefix matches
-- Injects only the **body content** of `SKILL.md` — the YAML front matter is stripped and never sent to the LLM
-- Skills directory is configured via `skills.dir` or `--skills-dir`
-
-**Role vs Skill**: A role defines the LLM's *personality*; a skill provides *task instructions* for how to carry out the user's request within that role. Skills are injected after the role and before the user prompt.
+- Returns the full `SKILL.md` content for inclusion in the prompt
 
 **Error Handling**:
+Errors are printed to stdout (visible to the user) and return `nil` so nothing is injected into the AI prompt:
 - Missing skill name: `Error: /skill requires a skill name. Use /skills to list available skills.`
-- No matching directory: `Error: No skill matching 'name' found in <dir>. Use /skills to list available skills.`
-- Directory exists but no SKILL.md: `Error: Skill directory 'name' has no SKILL.md. Use /skills to list available skills.`
-
-**See also**: `/skills` to list available skills, `--list-skills` CLI flag
+- No matching directory: `Error: No skill matching 'name' found in <skills_dir>. Use /skills to list available skills.`
+- Directory exists but no SKILL.md: `Error: Skill 'name' has no SKILL.md in <skills_dir>.`
 
 ### `/skills`
-List available AIA skills with optional search filtering.
+List available AIA skills, optionally filtered by search terms.
 
-**Syntax**: `/skills [search terms]`
+**Syntax**: `/skills [search_terms...]`
 
-**Examples**:
-```markdown
-/skills                   # List all skills
-/skills ruby              # Skills matching "ruby"
-/skills -python           # Skills not matching "python"
-/skills ruby -deprecated  # Skills matching "ruby" but not "deprecated"
+**Example Output** (`/skills`):
 ```
+code-assist: Code Assist
+  Help with coding tasks.
 
-**Search term prefixes**:
-- Bare word or `+word` — positive filter (skill must match)
-- `-word`, `~word`, or `!word` — negative filter (skill must not match)
-- Searches skill name and SKILL.md front matter fields (name, description, etc.)
+code-quality: Code Quality
+  Enforce SOLID principles and clean code guidelines across all
+  language targets.
+
+frontend-design: Frontend Design
+  Build interfaces.
 
-**Example Output**:
 ```
-Available Skills (~/.prompts/skills):
-======================================
 
-  code-quality
-    Improve code quality through comprehensive analysis and refactoring.
+**Search Term Filtering**:
+- `/skills ruby arch` — show only skills whose YAML front matter contains both `ruby` AND `arch` (AND logic, case-insensitive substring match)
+- Prefix `+` is treated as positive (same as bare term): `/skills +ruby`
+- Prefix `-`, `~`, or `!` is AND NOT — excludes matching skills: `/skills ruby -test`
+- Multiple negative terms each exclude independently
 
-  frontend-design
-    Create distinctive, production-grade frontend interfaces.
+**Features**:
+- Only lists subdirectories that contain a `SKILL.md` file; plain files and dirs without `SKILL.md` are ignored
+- Skills directory: `AIA.config.skills.dir` (default `~/.prompts/skills`); if `AIA_PROMPTS__SKILLS_PREFIX` is set, appended to `AIA_PROMPTS__DIR` to form the base
+- Results sorted alphabetically by skill ID
+- Descriptions are word-wrapped to terminal width with 2-space indent on continuation lines
+- Displays to STDOUT only (does not inject content into the AI prompt)
 
-Total: 2 skills
+## Execution Directives
+
+### `/ruby`
+Execute arbitrary Ruby code and insert the result.
+
+> **Security note:** `/ruby` executes the provided expression via `eval()` with no sandboxing. It has full access to the Ruby environment, the filesystem, and all loaded gems — equivalent to running arbitrary Ruby code in your terminal. Only use `/ruby` in prompts you trust. Do not share prompts containing `/ruby` unless you have reviewed their content.
+
+**Syntax**: `/ruby expression`
+
+**Examples**:
+```markdown
+/ruby Time.now.strftime('%Y-%m-%d')
+/ruby Dir.glob('*.rb').size
+/ruby ENV['USER']
 ```
 
 **Features**:
-- Lists skill subdirectories from `AIA.config.skills.dir` (default: `~/.prompts/skills/`)
-- Sorted alphabetically
-- Displays to STDOUT only (does not inject content into the prompt)
-- Reads `SKILL.md` front matter for filtering and description display
+- Evaluates the Ruby expression and inserts the string result into the prompt
+- Has full access to the Ruby environment and loaded gems
+- Returns error details if execution fails
 
-## Execution Directives
+**Aliases**: `/rb`
 
 ### `/say`
 Speak text using system text-to-speech (macOS/Linux).
@@ -239,6 +389,199 @@ Speak text using system text-to-speech (macOS/Linux).
 - macOS: Uses built-in `say` command
 - Linux: Requires `espeak` or similar TTS software
 
+### `/concurrent`
+Execute the next prompt with concurrent MCP server access. AIA groups independent MCP servers and queries them in parallel, then synthesizes the results.
+
+**Syntax**: `/concurrent`
+
+**Examples**:
+```markdown
+/concurrent
+What files changed this week and are there any open issues related to them?
+```
+
+**Features**:
+- Enables concurrent MCP mode for the next prompt only
+- Independent MCP servers are queried in parallel
+- A "Weaver" synthesizer robot merges the results
+- Requires at least 2 MCP servers to be connected
+
+**Aliases**: `/conc`
+
+### `/verify`
+Run the next prompt through a verification network. Two independent robots answer the same question, then a reconciler compares and merges the responses.
+
+**Syntax**: `/verify`
+
+**Examples**:
+```markdown
+/verify
+Is this SQL query safe from injection attacks?
+```
+
+**Features**:
+- Two robots independently answer the prompt
+- A third robot reconciles any differences
+- Increases confidence in factual or analytical answers
+- Requires a multi-model or network configuration
+
+### `/decompose`
+Decompose the next prompt into independent, concurrently-executable sub-tasks. The prompt is analyzed by a coordinator model, broken into pieces that have **no dependencies on each other**, each solved in parallel via `Async::Barrier`, and the results synthesized into a single response.
+
+**Syntax**: `/decompose`
+
+**Examples**:
+```markdown
+/decompose
+Analyze this codebase: review the test coverage, check for security issues, and assess performance.
+```
+
+**Features**:
+- Sub-tasks returned are **always guaranteed to be independent** — they can all run concurrently without waiting on each other
+- If the prompt cannot be split into parallel work (e.g., tasks that must run sequentially or build on each other's output), the model returns an **empty subtasks array** and the prompt executes normally as a single request
+- 2–5 sub-tasks maximum; single-step prompts are never decomposed
+- Results from all parallel branches are synthesized into one coherent final response by the coordinator
+- Best for prompts with multiple truly distinct aspects (e.g., "review X, analyze Y, and summarize Z" where X, Y, Z are independent)
+- **Not suitable for sequential workflows** — use `/pipeline` or `/next` when tasks must chain
+
+> **Design guarantee**: The decomposition model is explicitly instructed to return an empty `subtasks` array whenever any sub-task depends on the result of another. Parallelism is never approximated — it is either fully concurrent or the prompt runs as a whole.
+
+### `/debate`
+Enable multi-round debate between robots in the network. Robots argue different perspectives and converge toward a consensus answer.
+
+**Syntax**: `/debate`
+
+**Examples**:
+```markdown
+/debate
+Should we use microservices or a monolith for this project?
+```
+
+**Features**:
+- Runs up to 5 rounds of debate between network robots
+- Each robot sees all previous arguments before responding
+- Debate ends early when robots signal convergence (CONVERGED keyword)
+- Results include all debate rounds formatted as markdown
+- Writes debate history to network memory for context
+- Requires a multi-model network (2+ models)
+
+### `/delegate`
+Delegate subtasks to specific robots via a structured TrakFlow plan. The lead robot decomposes the prompt into a JSON task plan, and robots execute steps sequentially.
+
+**Syntax**: `/delegate`
+
+**Examples**:
+```markdown
+/delegate
+Build a REST API with authentication, rate limiting, and documentation.
+```
+
+**Features**:
+- Lead robot creates a structured task decomposition (JSON)
+- TaskCoordinator creates a TrakFlow plan from the decomposition
+- Each subtask is assigned to and executed by a specific robot
+- Prior step results flow as context into subsequent steps
+- Creates a full execution trace with per-step results
+- Requires a multi-model network
+
+**Aliases**: `/del`
+
+### `/spawn`
+Spawn a specialist robot for the **next prompt only**. The specialist is a
+one-shot helper — it does not join the crew. To add a persistent, `@mention`-able
+member instead, use [`/add_recruit`](#add_recruit).
+
+**Syntax**: `/spawn` | `/spawn <type>` | `/spawn <name> <provider/model> <system prompt...>`
+
+**Examples**:
+```markdown
+# Auto-detect the best specialist type from the prompt
+/spawn
+Write a comprehensive test suite for the User model.
+
+# Named specialist type
+/spawn security
+Audit this authentication module for vulnerabilities.
+
+# Fully explicit: name, provider/model, and system prompt
+/spawn researcher ollama/qwen3.6:latest You are careful and cite sources.
+Summarize the latest changes to the locking strategy.
+```
+
+**Features**:
+- No arguments: auto-detects the best specialist type from the prompt content.
+- A single `<type>` argument: spawns a specialist of that type (e.g., `security`, `testing`, `performance`), inheriting the current model with a domain-specific system prompt.
+- Explicit form (`<name> <provider/model> <system prompt>`): sets the name, model/provider, and system prompt directly. Use `-` or `inherit` as the model token to keep the parent's model; `lms/...` maps to the `openai` provider.
+- Specialist robots are cached and reused within the session.
+- Creates a TrakFlow task when task coordination is active.
+
+### `/add_recruit`
+Recruit a new robot into the crew. Unlike `/spawn`, a recruit **persists** for
+the rest of the session, appears in [`/robots`](#robots), and answers to
+`@name`. Recruits inherit the chief's local tools and connected MCP servers.
+
+**Aliases**: `/add`
+
+**Syntax**: `/add_recruit <name> [provider/model] [skill:<id>...] [system prompt...]`
+
+**Examples**:
+```markdown
+# Inherit the chief's model and a default persona
+/add_recruit researcher
+
+# Explicit provider/model plus a system prompt
+/add_recruit critic anthropic/claude-3-5-sonnet You are a ruthless design critic.
+
+# A local-model member
+/add_recruit local ollama/qwen3.6:latest You are concise and fast.
+
+# Assign a skill as the member's role (divides labor across the crew)
+/add_recruit reviewer skill:security-review
+/add_recruit reviewer - skill:security-review,ruby-style focus on the auth module
+```
+
+**Features**:
+- `<name>` is required and must be unique within the crew. `crew` is reserved (it is the broadcast handle, `@crew`).
+- Omit `provider/model` to inherit the chief's model and provider; use `-` or `inherit` to inherit explicitly while still supplying skills or a system prompt.
+- A `skill:<id>` token (anywhere after the name; comma-separate or repeat for several) assigns one or more [skills](guides/crew.md#giving-a-recruit-a-role-with-skills) as the member's role. The skill bodies become its system prompt; an unknown id is reported.
+- Everything else after the model becomes the member's system prompt, appended after any skills.
+- Address the recruit afterward with `@name`, or broadcast to the whole crew with `@crew`.
+
+### `/reskill`
+Reset a crew member to a clean slate (a fresh conversation) and give it a new
+role. The member keeps its `@name` and model. The chief cannot be reskilled.
+
+**Syntax**: `/reskill <name> [skill:<id>...] [system prompt...]`
+
+**Examples**:
+```markdown
+# Repurpose a member with a different skill
+/reskill larry skill:test-writing
+
+# Or hand it a freeform role
+/reskill larry you now summarize the other members' findings
+```
+
+**Features**:
+- Same `skill:<id>` and trailing-system-prompt handling as `/add_recruit` (there is no model token — the member's existing model is preserved).
+- Useful for recovering a member that drifted off task, or moving the crew through phases (review → fix → summarize) without re-creating robots.
+
+### `/drop_recruit`
+Remove a robot from the crew by name. The chief (the session's lead robot)
+cannot be dropped.
+
+**Aliases**: `/drop`
+
+**Syntax**: `/drop_recruit <name>`
+
+**Examples**:
+```markdown
+/drop_recruit researcher
+```
+
+See the [Crews guide](guides/crew.md) for `@mention` routing, `@crew`
+broadcast, and concurrent vs. sequential execution.
+
 ## Utility Directives
 
 ### `/tools`
@@ -288,50 +631,156 @@ FileReader
 - Filtering is case-insensitive (e.g., "File", "FILE", and "file" all match)
 
 ### `/next`
-Set the next prompt to execute in a workflow.
+Declare the next prompt to execute after this one in a workflow.
+
+> **Note**: `next` is a **YAML front matter key** written at the top of a `.md` prompt
+> file, not a chat-time slash command. It is processed by the prompt loader when the
+> file is read, not interactively during a chat session.
+
+**Syntax** (YAML front matter in a prompt file):
+```yaml
+---
+next: analyze_results
+---
+Your prompt body here.
+```
 
-**Syntax**: `/next prompt_id`
+**Examples** (prompt files with `next` front matter):
+```yaml
+---
+next: generate_report
+---
+Summarize the extracted data.
+```
 
-**Examples**:
-```markdown
-/next analyze_results
-/next generate_report
+```yaml
+---
+next: finalize
+---
+Analyze the results and identify key findings.
 ```
 
-**Usage**:
-- `/next` - Display current next prompt
-- `/next prompt_id` - Set next prompt in workflow
+**Usage**: When AIA finishes running a prompt that has a `next` key in its front
+matter, it automatically loads and executes the named prompt next.
 
 ### `/pipeline`
-Define or modify a prompt workflow sequence.
+Declare an ordered sequence of prompts to run as a workflow.
+
+> **Note**: `pipeline` is a **YAML front matter key** written at the top of a `.md`
+> prompt file, not a chat-time slash command. It is processed by the prompt loader
+> when the file is read, not interactively during a chat session. The alias
+> `workflow` is also accepted as a front matter key.
+
+**Syntax** (YAML front matter in a prompt file):
+```yaml
+---
+pipeline: prompt1, prompt2, prompt3
+---
+Your prompt body here.
+```
 
-**Syntax**: `/pipeline prompt1,prompt2,prompt3`
+**Examples** (prompt files with `pipeline` front matter):
+```yaml
+---
+pipeline: extract_data, analyze, report
+---
+Begin the automated data processing workflow.
+```
 
-**Examples**:
-```markdown
-/pipeline extract_data,analyze,report
-/pipeline code_review,optimize,test
+```yaml
+---
+workflow: code_review, optimize, test
+---
+Start the code quality pipeline.
 ```
 
-**Usage**:
-- `/pipeline` - Display current pipeline
-- `/pipeline prompts` - Set pipeline sequence
-- Can use comma-separated or space-separated prompt IDs
+**Usage**: When AIA loads a prompt that has a `pipeline` (or `workflow`) key in its
+front matter, it queues all listed prompts in order and runs them sequentially.
 
-**Aliases**: `/workflow`
+**Aliases**: `workflow` (as a front matter key)
 
-### `/terse`
-Add instruction for brief responses.
+### `/mcp`
+Show MCP server connection status and available tools per server.
 
-**Syntax**: `/terse`
+**Syntax**: `/mcp`
 
-**Example**:
-```markdown
-/terse
-Explain machine learning algorithms.
+**Example Output**:
+```
+MCP Server Status
+=================
+Defined: 5  Connected: 3  Failed: 2
+
+Connected Servers:
+  github (12 tools)
+    - github_create_issue
+    - github_list_repos
+    ...
+  filesystem (4 tools)
+    - read_file
+    - write_file
+    ...
+
+Failed Servers:
+  slack: Connection refused
+  jira: Authentication failed
 ```
 
-Adds: "Keep your response short and to the point." to the prompt.
+**Features**:
+- Shows count of defined, connected, and failed MCP servers
+- Lists each connected server with its tool count and tool names
+- Shows error details for failed server connections
+- Respects `--mcp-use` and `--mcp-skip` filtering
+
+### `/robots`
+Show active robot configuration — the current robot or network topology.
+
+**Syntax**: `/robots`
+
+**Example Output** (single robot):
+```
+Active Robot
+============
+Mode: Single
+
+  Tobor
+    Model:    gpt-4o (openai)
+    Wage:     $0.0050 in / $0.0150 out per 1K tokens
+    Tools:    8 (3 local, 5 mcp)
+    Role:     (default)
+```
+
+**Example Output** (multi-model network):
+```
+Active Robots
+=============
+Mode: Consensus Network (3 robots)
+
+  Tobor
+    Model:    gpt-4o (openai)
+    Wage:     $0.0050 in / $0.0150 out per 1K tokens
+    Tools:    5 (2 local, 3 mcp)
+    Role:     architect
+
+  Spark
+    Model:    claude-sonnet-4-20250514 (anthropic)
+    Wage:     $0.0030 in / $0.0150 out per 1K tokens
+    Tools:    5 (2 local, 3 mcp)
+    Role:     security
+
+  Weaver
+    Model:    gpt-4o (openai)
+    Wage:     $0.0050 in / $0.0150 out per 1K tokens
+    Tools:    0
+    Role:     (default)
+```
+
+**Features**:
+- Shows mode: Single, Parallel, Consensus, or Pipeline
+- Displays robot name, model, provider, cost, tools, and role
+- For networks, shows all robots including the Weaver synthesizer
+
+### `/terse`
+*Deprecated.* Previously added an instruction for concise responses. Now a no-op.
 
 ### `/robot`
 Generate ASCII art robot.
@@ -364,7 +813,7 @@ Create a named checkpoint of the current conversation context.
 - `/checkpoint` - Create an auto-named checkpoint
 - `/checkpoint name` - Create a checkpoint with specific name
 
-**Aliases**: `/cp`
+**Aliases**: `/ckp`, `/cp`
 
 ### `/restore`
 Restore conversation context to a previously saved checkpoint.
@@ -430,6 +879,72 @@ Checkpoints: ruby_basics, oop_concepts
 - Truncates long messages for readability (200 characters)
 - Shows total message count and checkpoint summary
 
+### `/checkpoints_list`
+List all saved checkpoints with their positions and timestamps.
+
+**Syntax**: `/checkpoints_list`
+
+**Example Output**:
+```
+=== Available Checkpoints ===
+  1: position 4, created 14:32:07
+    → "Tell me about Ruby programming"
+  before_refactor: position 8, created 14:45:22
+    → "Now refactor the User model"
+=== End of Checkpoints ===
+```
+
+**Features**:
+- Shows checkpoint name, conversation position, and creation time
+- Includes a preview of the last user message at the checkpoint
+- Displays "No checkpoints available." when none exist
+
+**Aliases**: `/checkpoints`
+
+## TrakFlow Directives
+
+TrakFlow is AIA's built-in task tracking system for managing multi-step work.
+
+### `/tasks`
+Show TrakFlow ready tasks and project summary.
+
+**Syntax**: `/tasks [summary]`
+
+**Examples**:
+```markdown
+/tasks              # Show ready tasks
+/tasks summary      # Show project summary
+```
+
+**Features**:
+- Without arguments: lists tasks that are ready to be worked on
+- With `summary`: shows the overall project status
+- Returns an initialization message if TrakFlow is not set up
+
+**Aliases**: `/tf`
+
+### `/plan`
+Create a TrakFlow plan from a description. The description is decomposed into actionable tasks.
+
+**Syntax**: `/plan description`
+
+**Examples**:
+```markdown
+/plan Build user authentication with login, signup, and password reset
+/plan Refactor the payment module into separate service objects
+```
+
+### `/task`
+Create a single TrakFlow task.
+
+**Syntax**: `/task title`
+
+**Examples**:
+```markdown
+/task Add input validation to the registration form
+/task Write integration tests for the API endpoints
+```
+
 ## Model and Information Directives
 
 ### `/available_models`