Slash Commands
Slash commands provide quick access to Pando features without leaving the chat interface. Type / at the start of an empty input to see available commands.
How to Use Slash Commands
TUI (Terminal User Interface)
Type / in the chat input. A fuzzy-searchable dialog appears showing all available commands organized by category. Use arrow keys or type to filter, then press Enter to select.
Web UI
Type / in the chat input. Commands appear in an autocomplete dropdown that filters as you type.
ACP (Editor Integration)
Type / in the editor chat panel. Commands are available in VS Code, Zed, and JetBrains through the Agent Client Protocol.
CLI (Non-Interactive)
Use flags instead of slash commands:
pando --goal "Fix tests" # equivalent to /goal
pando --goal "Refactor auth" --model copilot.gpt-5.4Goal Mode Commands
| Command | Description |
|---|---|
/goal <objective> | Start autonomous goal mode with a persistent objective |
/autopilot <objective> | Alias for /goal |
/goal-status | Show the status of the current goal (progress, iterations, elapsed time) |
/goal-cancel | Cancel the running goal |
Goal mode lets Pando work autonomously across multiple turns without requiring user input after each step. See Goal Mode for full details.
Session Management Commands
| Command | Description |
|---|---|
/compact | Create a manual compact summary for the current session (frees tokens) |
/summarize | Alias for /compact |
/db-compact | Compact the database (SQLite VACUUM) to reclaim free space |
/compact summarizes the conversation history to free up context window tokens. Use this when the session is getting long and you want to continue working without starting a new session.
/db-compact runs SQLite VACUUM on Pando’s database to reclaim disk space from deleted records. This is useful after长时间使用 or after cleaning up old sessions.
Code Quality Commands
Ponytail Mode
Ponytail is a “lazy senior developer” mode that applies YAGNI (You Aren’t Gonna Need It) principles to code suggestions.
| Command | Description |
|---|---|
/ponytail lite | Enable lite YAGNI mode - suggests simpler alternatives |
/ponytail full | Enable full YAGNI mode - more aggressive about simplicity |
/ponytail ultra | Enable ultra YAGNI mode - maximum simplicity enforcement |
/ponytail off | Disable ponytail mode |
Caveman Mode
Caveman reduces output verbosity to save output tokens while keeping code, commands, and errors exact.
| Command | Description |
|---|---|
/caveman lite | Drop filler and restatement. Normal sentences, fewer of them |
/caveman full | Terse by default. Conclusions and fragments, no unrequested explanation |
/caveman ultra | Maximum brevity. The answer and nothing around it |
/caveman-finish | Disable caveman and return to normal output |
Workflow Commands
Superpowers Mode
Superpowers enforces a disciplined development workflow: plan-first, verify-always.
| Command | Description |
|---|---|
/superpowers [objective] | Enable the disciplined development workflow |
/superpowers-finish | Verify, report, and return to normal mode |
When active, work follows this lifecycle:
- Understand before designing
- Design, then get approval
- Plan long work explicitly
- Implement test-first, in small increments
- Verify with evidence, not claims
- Review before declaring ready
Learning Mode
Learning mode makes Pando work as a deliberate learner and documentarian.
| Command | Description |
|---|---|
/learning [focus] | Enable learner mode: read KB more, document discoveries, ask questions |
/learning-finish | Consolidate learnings into KB/memory and return to normal mode |
When active, Pando:
- Searches the knowledge base before acting
- Asks the user instead of guessing
- Captures non-trivial discoveries as documentation
- Keeps existing docs honest (updates stale, marks superseded)
Project Management Commands
| Command | Description |
|---|---|
/improve-agents-md [guidance] | Create or reinforce AGENTS.md with mandatory AI-agent operating rules |
Security Commands
Vulnerability Hunter
Adversarial security audit workflow ported from Capital One’s VulnHunter. These commands run self-contained security workflows as normal agent turns — they stream, steer, and persist like any other message.
| Command | Description |
|---|---|
/vulnhunt [scope] | Trace attacker input to sinks and report exploitable vulnerabilities |
/vulnhunter-fix [finding] | Test-driven remediation: exploit → failing test → fix → verify |
/vulnhunt-fix-verify [findings] | Read-only independent verification of claimed security fixes |
/vulnhunt runs a full security audit: recon → parallel class-group hunt → exploitability verify → adversarial disprove → capability-filtered report. Findings are persisted to the knowledge base. Pass an optional scope (subdirectory, package, or emphasis) to narrow the audit.
/vulnhunter-fix performs test-driven remediation of confirmed vulnerabilities: creates an exploit proof, writes a failing security test (RED), implements the fix (GREEN), and verifies the exploit is blocked without regressions.
/vulnhunt-fix-verify is a read-only verification that checks claimed fixes against the actual code, emitting a per-finding verdict: FIXED, PARTIAL, NOT_FIXED, or INCONCLUSIVE.
Custom Commands
Pando supports custom slash commands via markdown files. Place .md files in:
- Project commands:
<data-dir>/commands/(shown asproject:command-name) - User commands:
~/.config/pando/commands/or~/.pando/commands/(shown asuser:command-name)
Custom commands appear in the command palette alongside built-in commands.
Command Categories
Commands are organized into categories in the command palette:
- General: Goal, compact, db-compact
- Code Quality: Ponytail and Caveman modes
- Workflow: Superpowers and Learning modes
- Project: AGENTS.md management
- Security: Vulnerability Hunter (vulnhunt, vulnhunter-fix, vulnhunt-fix-verify)