Skip to main content

​
All Workflows

Workflows are slash commands that trigger predefined sequences of actions. They’re the backbone of Athena’s session management and deep reasoning capabilities.

​
Quick Reference

CommandPurposeComplexity
/startBoot session, load identityLow
/endClose session, commit to memoryLow
/saveMid-session checkpointLow
/thinkDeep reasoning (all phases)Medium
/ultrathinkMaximum depth (parallel orchestrator)High
/researchExhaustive multi-source investigationHigh
/searchWeb search with citationsMedium
/planStructured planning with pre-mortemMedium
/briefPre-prompt clarification protocolMedium
/refactorFull workspace optimizationHigh
/vibeShip at 70%, iterate fastLow
/deploySanitized public repo syncMedium

​
Core Workflows by Category

​
Session Management

​
/start

When to Use: Beginning of every session Loads identity, recalls context from previous sessions. Ultra-low latency boot (less than 2K tokens) following the “boot fast, load later” philosophy. View details →

​
/end

When to Use: End of every session Commits insights to memory, updates indexes. Reviews session, synthesizes key decisions, and creates git commit. View details →

​
/save

When to Use: Mid-session checkpoint Use before risky experiments or when switching contexts. Creates lightweight checkpoint without full maintenance overhead. View details →

​
/tutorial

When to Use: Your very first session Guided walkthrough of all features, includes profile interview. Best for onboarding new users.

​
Reasoning Depth

​
/think

Depth: High | Token Budget: ~2000 tokens When to Use: Important decisions, complex problems Executes all phases (0-VII) including counterfactuals and confrontation. Mandatory stress-testing and multi-path reasoning. View details →

​
/ultrathink

Depth: Maximum | Token Budget: ~5000+ tokens When to Use: Life-altering decisions, multi-stakeholder analysis Parallel orchestrator mode (v3.0) with true parallel API calls, adversarial convergence gate, and recursive analysis. Includes Senior Principal Review (Protocol 335) for architecture and technical decisions. Rule: Default to normal mode. Escalate to /think for $10K+ decisions. Use /ultrathink only for maximum-depth analysis.

​
/search

Searches: 2-3 | Sources Read: 0-2 | Depth: Medium When to Use: Fact-checking, current events, recent developments Multi-query search with cross-referencing, citation-dense output, and blind-spot coverage.

​
/research

Searches: 5-10+ | Sources Read: 3-10+ | Depth: Maximum When to Use: Market research, regulatory deep-dives, due diligence Exhaustive investigation with 6 mandatory layers:
  1. Initial survey (5-10 searches)
  2. Source deep-dive
  3. Follow rabbit holes (2-3 levels)
  4. Cross-domain synthesis
  5. Adversarial stress-test
  6. Actionable distillation
View details →

​
Triple Crown Mode (/think /search /research)

When to Use: Life-altering decisions, maximum stakes Combines all three for nuclear-level investigation:
  • 10-20+ exhaustive searches
  • 5-10 full articles read
  • 3+ level rabbit holes
  • Full Phase 0-VII analysis
  • Adversarial red-team testing
  • Permanent Codex deposit

​
Planning Workflows

​
/brief — Pre-Prompt Clarification

When to Use: Before complex tasks to clarify requirements Measure twice, cut once. Clarifies requirements before wasting tokens on wrong output. Variants:
  • /brief — Core brief (default)
  • /brief ++ — Expanded fields for complex work
  • /brief build — Technical implementation tasks
  • /brief research — Investigation/analysis tasks
  • /brief interview — Iterative Q&A to extract requirements
Interview Mode (v3.1): Iterative questioning (10 max per turn) instead of overwhelming dumps. Continues until full clarity is achieved, then writes spec to .context/specs/[PROJECT]_SPEC.md.

​
/plan — Structured Task Planning

When to Use: Heavy tasks (new features, refactors, architecture changes) Process:
  1. Enter PLANNING mode
  2. Generate implementation plan with pre-mortem
  3. Review with user before execution
  4. Track progress in task.md
Applies Protocol 272 (Harness Engineering): Define constraints so tightly that execution is trivial.

​
Maintenance Workflows

​
/refactor — Full Workspace Optimization

Latency: HIGH (~15-30 min) | Philosophy: Flawless or nothing When to Use: Periodically to maintain workspace integrity Phases:
  1. Diagnostics (invoke /diagnose)
  2. Pre-remediation checkpoint
  3. Fix orphans/broken links
  4. Optimization pass
  5. Supabase memory sync
  6. Cache refresh
  7. Index regeneration
  8. Regression test validation
  9. Commit
Flags: --dry-run to preview changes Refactor Levels:
  • Level 1 (Hygiene): < 1 week since last, formatting only
  • Level 2 (Component): Modules broken, schemas drifted
  • Level 3 (Architecture): Core framework rot, perf bottlenecks
  • Level 4 (Rewrite): “Red Zone” (Avoid)

​
/vibe — Vibe Coding Mode

Philosophy: Speed > Perfection. Ship at 70%. When to Use: Rapid iteration, prototyping Principles:
  • Prototype in hours, not days
  • Ship → Feedback → Improve
  • Working > Perfect
  • Bias to action
SEO Pre-Ship Checklist (mandatory for pages):
  • Keyword in H1/Title
  • Internal linking
  • Descriptive URL slug
  • Meta description

​
Business & Strategy Workflows

​
/due-diligence (alias: /dd)

When to Use: Friend asks for money, startup pitch, franchise opportunity Process:
  1. Kill Switch Check (5 min) - Regulatory/structural deal-breakers
  2. Framework Cascade (15 min) - PESTLE, Five Forces, SWOT, Funding Source Ladder
  3. Financial Model (15 min) - 3 scenarios (Worst/Realistic/Best) with EV calculation
  4. Human-AI Discussion (10 min) - Clarify assumptions
  5. Cross-Model Peer Review (15 min) - Submit to multiple SOTA models
  6. Final Synthesis (10 min) - Report with recommendation
Red Flags:
  • Approaching friends, not banks (institutional rejection likely)
  • No personal capital at risk (zero skin in game)
  • Undefined exit mechanism (no liquidity)
  • “Trust me” instead of data (governance risk)

​
/brand-generator — Agency Arbitrage

Value: 10,000(AgencyRate)vs 10,000 (Agency Rate) vs ~0 (AI Rate) | Time: ~20 Minutes When to Use: Building a brand or refreshing existing brand identity Phases:
  1. Brand DNA Extraction (Optional) - Use Google Labs (Pomelli) to extract from existing URL
  2. Strategic DNA - Generate Mission, Values, Voice, Archetype
  3. Visual Synthesis - Generate high-fidelity logo concepts
  4. Asset Refining - Convert to professional vector assets
  5. Brand Book - Package into deliverable PDF
Tools: Athena + Vectorizer.ai + Gamma.app

​
Deployment & Sync Workflows

​
/deploy — Public Repo Synchronization

When to Use: Publishing sanitized content to public repository Process:
  1. Context Assessment - Identify what’s ready for public release
  2. Sanitization Protocol - Strip PII, financials, locations, harsh tone
  3. Deployment Execution - Copy to public repo structure
  4. Git Synchronization - Commit and push
  5. Verification - Confirm success
Explicit Exclusions:
  • Psychology Profiles
  • Risk Playbooks
  • Private Case Studies (unless fully sanitized)

​
Creating Custom Workflows

Workflows live in .agent/workflows/ as Markdown files: 
---
description: Short description for the workflow index
created: YYYY-MM-DD
last_updated: YYYY-MM-DD
---

# /command-name — Title

## Behavior

What happens when invoked...

## Phases

Step-by-step execution...

## Tagging

#workflow #automation
Conventions:
  • Use // turbo annotation above steps that are safe to auto-run
  • Reference other workflows with relative links
  • Include rollback instructions for destructive operations

​
Best Practices

  1. Choose the right depth: Don’t /ultrathink on simple queries
  2. Checkpoint often: Use /save before risky experiments
  3. Brief first: For complex tasks, /brief reduces wasted tokens
  4. End sessions properly: /end commits insights to long-term memory
  5. Combine strategically: /think /search /research for maximum coverage

​
Workflow Comparison Matrix

WorkflowLatencyToken BudgetAuto-RunDestructiveOutput
/startUltra-LowLess than 2KPartialNoSession init
/endLowLess than 1KPartialNoGit commit
/saveUltra-LowLess than 500YesNoCheckpoint
/thinkMedium~2KNoNoAnalysis
/ultrathinkHigh~5K+NoNoDeep analysis
/searchLowLess than 1KNoNoCited findings
/researchHigh~3K+NoNoResearch report
/briefLowLess than 1KNoNoSpec document
/planMedium~1KNoNoImplementation plan
/refactorVery High~2KPartialYesWorkspace optimization
/vibeVariableVariableNoNoRapid prototype
/deployMediumLess than 1KPartialYesPublic sync
/due-diligenceHigh~2KNoNoInvestment report
/brand-generatorMedium~1KNoNoBrand package

​
Further Reading

Build docs developers (and LLMs) love

Get started for freeTalk to us