Slash Commands Reference

This is the complete reference for all slash commands available inside kognisant chat. Commands are organized by category for quick lookup.

General

/help

Display the full command reference matrix.

code

/help

For detailed help on a specific command:

code

/help agent

/clear

Flush the current conversation history. Starts fresh while preserving the system prompt and project context.

code

/clear

This does NOT clear persistent memory (context.md). It only resets the rolling conversation window.

/paste (or /p)

Enter multi-line paste mode. Useful for pasting large log traces, error messages, or code blocks.

code

/paste

Type or paste your content, then type /end on a blank line to submit:

code

/paste
ERROR: Connection refused at line 42
Traceback:
  File "server.py", line 42
    socket.connect(...)
ConnectionRefusedError: [Errno 111]
/end

---

Context and Files

/context

Display the project's persistent memory (context.md). Shows milestones, active tasks, architecture notes, and anything the AI has learned about your project.

code

/context

/files

List all indexed files in your project workspace (excludes patterns like .git, node_modules, __pycache__, virtual environments).

code

/files

/read

Load a file's full content into the conversation context. The AI can then reference, explain, or modify it.

code

/read src/auth/middleware.py

code

/read tests/test_auth.py

You can read multiple files by calling /read multiple times. Each file is added to the conversation history.

---

Model Management

/model

Open the interactive model selection menu. Switch between configured models, add a new one, or remove an existing one.

code

/model

Shows:

code

📦 Select an AI Model:

  [1] gemma3:1b (Ollama) 🟢
  [2] llama-3.3-70b-versatile (Groq) 🟢 [Active]
  [3] gpt-4o-mini (OpenAI) 🟡
  [a] Add custom provider / model
  [r] Remove a model from pool

Options:

Select a number to switch instantly

Press a to add a new provider

Press r to remove a model (cannot remove the active model)

/providers

Inspect configured AI providers with pricing, API key status, and endpoint details.

code

/providers

---

Agents

/agent

Dispatch an autonomous PERP swarm to handle a complex multi-step task.

code

/agent Write a complete test suite for the database module

code

/agent Research OAuth2 PKCE and implement it in our auth system

The swarm plans, executes in parallel, reflects, and persists results.

/status

Check the current state of running agents. Shows active workers, their subtasks, and progress.

code

/status

/stop

Cancel the running agent swarm. Currently active workers finish their task, but no new subtasks are dispatched.

code

/stop

/pause

Suspend agent dispatch. Workers in progress continue, but no new subtasks start.

code

/pause

/resume

Continue a paused agent swarm. Remaining subtasks begin dispatching again.

code

/resume

---

Telemetry and Monitoring

/telemetry

Show execution statistics summary (last 50 executions).

code

/telemetry

For per-model deep dive:

code

/telemetry gemma4:latest

Shows success rate, response times, reliability score, failure breakdown, and circuit breaker state.

/thinking

Review AI reasoning/thinking steps from past executions.

code

/thinking last Most recent thinking

/thinking 3 Thinking from 3 executions ago

/thinking list List all stored thinking records

---

World Model and Goals

/worldmodel

Manage the World Model subsystem.

code

/worldmodel enable Enable codebase analysis

/worldmodel disable Disable (data preserved)

/worldmodel status Show state and statistics

/goals

View active improvement goals generated by the world model.

code

/goals

Shows ranked goals with priority:

code

Goals:
  [1] ⚠️  Fix contract violation: auth → jwt_utils (priority: HIGH)
  [2] 📊 Add tests for connection_pool.py (priority: MEDIUM)
  [3] 🔄 Review legacy_parser.py (priority: LOW)

/goals accept

Accept a goal for execution. The system executes it immediately or queues it depending on graduated autonomy settings.

code

/goals accept 1

/goals dismiss

Dismiss a goal. Records negative feedback so the system learns your preferences.

code

/goals dismiss 3

---

Background Daemon and Jobs

/daemon status

Show daemon running state, PID, and uptime.

code

/daemon status

/daemon start

Start the background daemon from within chat.

code

/daemon start

/daemon stop

Stop the running daemon (sends SIGTERM for graceful shutdown).

code

/daemon stop

/daemon restart

Stop and restart the daemon in one command.

code

/daemon restart

/jobs

List all background jobs with name, type, state, run count, exit code, next scheduled run, and PID.

code

/jobs

Output:

code

NAME TYPE STATE RUN# EXIT LAST RUN NEXT RUN

telegram-bot persistent running 3 - 2025-06-15T10:30 UTC -

nightly-tests scheduled scheduled 12 0 2025-06-15T02:00 UTC in 8h 30m

/job stop

Cancel a running job. Sends SIGTERM to the subprocess and sets state to "cancelled".

code

/job stop telegram-bot

/job logs

View the last 30 lines of a job's output log.

code

/job logs telegram-bot

/job restart

Restart a stopped or crash-looped persistent job.

code

/job restart telegram-bot

/job remove

Permanently remove a job from the queue. Terminates the subprocess if currently running.

code

/job remove old-bot

---

Spec-Driven Development

/spec

Manage feature specifications from within chat.

code

/spec list List all specs with status

/spec auth_module Load spec context into conversation

/spec auth_module run Execute the next unchecked task

/spec auth_module run all Execute all remaining tasks

/spec auth_module done Mark the current task as complete

---

Skills and Tools

/skills

Display all loaded global skills (markdown knowledge files).

code

/skills

/tool list

List all registered global tools with their descriptions.

code

/tool list

/tool register

code

/tool register my_tool scripts/my_tool.py

/tool delete

Remove a global tool (deletes both JSON schema and Python implementation).

code

/tool delete my_tool

---

Session Control

exit / quit

End the chat session. Saves conversation history to .kognisant/history/ and exits cleanly.

code

exit

code

quit

---

Quick Reference Table

Command	Arguments	Description
:---	:---	:---
`/help`	`[command]`	Show help (optionally for a specific command)
`/clear`		Reset conversation history
`/paste`		Enter multi-line input mode
`/context`		View project memory
`/files`		List project files
`/read`		Load file into context
`/model`		Switch, add, or remove models
`/providers`		View provider details
`/agent`		Deploy autonomous swarm
`/status`		Check agent progress
`/stop`		Cancel running agent
`/pause`		Pause agent dispatch
`/resume`		Resume paused agent
`/telemetry`	`[model]`	Execution statistics
`/thinking`	`last/N/list`	Review reasoning steps
`/worldmodel`	`enable/disable/status`	Manage world model
`/goals`	`[accept/dismiss N]`	View or act on goals
`/daemon`	`start/stop/restart/status`	Manage background daemon
`/jobs`		List all jobs
`/job stop`		Cancel a job
`/job logs`		View job output
`/job restart`		Restart a job
`/job remove`		Delete a job
`/spec`	`[name] [action]`	Spec-driven development
`/skills`		List loaded skills
`/tool`	`list/register/delete`	Manage global tools
`exit`		End session

Slash Commands Architecture