Workflow: Timetravel Strategy Create

Design and implement a custom search strategy in ~/digital/leviathan/plugins/timetravel/.

Every step spawns a fresh subagent (poly sparkle pattern). State flows only through handoff artifacts at .lev/pm/handoffs/tt-strategy-create-*.md.

Trigger

• •User needs a custom adapter mix for a specific domain or use case
• •Existing strategies (quick, full, deep, max, academic, social) don't fit
• •/workflow tt-strategy-create "<name>" "<description>" [--adapters a,b,c]

Inputs

Steps

Step 1: Research Use Case

Agent Type: Explore Skills: skill://lev-research Research:

• •Analyze the strategy description for: domain, depth requirements, speed requirements, source types needed
• •Read existing strategies: ~/digital/leviathan/plugins/timetravel/src/strategy.ts
• •Read available adapters: ~/digital/leviathan/plugins/timetravel/src/adapters/index.ts
• •Check adapter capabilities vs use case needs

Action:

• •Classify use case: speed-optimized, depth-optimized, breadth-optimized, domain-specific
• •Identify which source types matter: web, academic, social, news, code, legal, etc.
• •Map source types to available adapters
• •Determine if synthesis is needed (complex/comparative queries = yes)

Handoff: .lev/pm/handoffs/tt-strategy-create-1-research.md

strategy_name: <name>
use_case_type: speed | depth | breadth | domain
source_types_needed: [web, academic, social, ...]
candidate_adapters: [<list with rationale>]
synthesis_needed: true | false
speed_requirement: fast | moderate | slow-ok

Step 2: Audit Existing Strategies

Agent Type: Explore Skills: None Research:

• •Read ~/digital/leviathan/plugins/timetravel/src/strategy.ts fully
• •Map each existing strategy: name, adapters, mode, synthesizer
• •Identify gaps: what combinations are NOT covered

Action:

• •Compare new strategy needs against existing 6 strategies
• •Check for overlap: if >80% same adapters as existing, recommend extending instead
• •Identify unique value proposition of new strategy
• •Verify no name collision

Handoff: .lev/pm/handoffs/tt-strategy-create-2-audit.md

existing_strategies:
  - name: quick
    adapters: [brave, tavily]
    mode: first-success
  - name: full
    adapters: [brave, firecrawl, perplexity]
    mode: parallel
  # ... all 6
overlap_with: <closest existing strategy or "none">
overlap_percentage: <0-100>
unique_value: "<what this strategy adds>"
proceed: true | false

Step 3: Design Adapter Mix

Agent Type: general-purpose Skills: None Research:

• •Read handoffs from Steps 1-2
• •Run cd ~/digital/leviathan/plugins/timetravel && npx timetravel status for current availability

Action:

• •Select 2-5 adapters with explicit justification per adapter
• •Consider: availability (prefer adapters with keys set), complementarity (diverse source types), reliability
• •If user provided --adapters override, validate they exist and document rationale
• •Define fallback behavior: what if primary adapters unavailable

Handoff: .lev/pm/handoffs/tt-strategy-create-3-design.md

selected_adapters:
  - name: brave
    role: "primary web search, fast baseline"
  - name: exa
    role: "neural/semantic for deeper relevance"
  - name: perplexity
    role: "AI-synthesized supplementary"
fallback_adapters: [<optional fallback list>]
adapter_count: <n>

Step 4: Define Orchestration

Agent Type: general-purpose Skills: None Research:

• •Read handoff from Step 3
• •Read orchestration modes in ~/digital/leviathan/plugins/timetravel/src/strategy.ts (executeStrategy)

Action:

• •Choose mode based on use case:
  • •parallel — best for breadth, depth (query all at once)
  • •sequential — best for cascading refinement (each builds on previous)
  • •first-success — best for speed (stop at first good result)
• •Choose synthesizer if needed (typically oracle for complex queries)
• •Document the full Strategy object shape

Handoff: .lev/pm/handoffs/tt-strategy-create-4-orchestration.md

mode: parallel | sequential | first-success
synthesizer: oracle | null
strategy_object:
  name: <name>
  adapters: [<list>]
  mode: <mode>
  synthesizer: <name or null>
rationale: "<why this orchestration>"

Step 5: Implement

Agent Type: general-purpose Skills: None Research:

• •Read handoff from Step 4
• •Read ~/digital/leviathan/plugins/timetravel/src/strategy.ts for exact insertion point

Action:

• •Add new strategy to the STRATEGIES constant in strategy.ts
• •Follow exact pattern of existing entries
• •Ensure TypeScript types are satisfied

Handoff: .lev/pm/handoffs/tt-strategy-create-5-implement.md

file_modified: src/strategy.ts
strategy_added: <name>
line_inserted_at: <approximate line>

Step 6: Test

Agent Type: general-purpose Skills: None Research:

• •Read handoff from Step 5

Action:

• •Run cd ~/digital/leviathan/plugins/timetravel && pnpm build — verify compilation
• •Run cd ~/digital/leviathan/plugins/timetravel && pnpm test — verify no regressions
• •Run npx timetravel search "test query for <strategy_name>" --strategy <strategy_name> — verify execution
• •Record: adapters used, source count, duration, synthesis presence

Handoff: .lev/pm/handoffs/tt-strategy-create-6-test.md

build_result: pass | fail
test_result: pass | fail
live_test:
  strategy: <name>
  adapters_responded: [<list>]
  source_count: <n>
  duration_ms: <n>
  has_synthesis: true | false
  status: pass | fail

Step 7: Register and Document

Agent Type: general-purpose Skills: None Research:

• •Read all handoffs from Steps 1-6
• •Read ~/.agents/skills/lev-research/SKILL.md for strategy mapping table

Action:

• •Update lev-research SKILL.md: add row to Strategy Mapping table
• •Update lev-research SKILL.md: add row to CLI Usage section
• •Update README.md if it exists
• •Update any related skill files that reference strategy lists

Handoff: .lev/pm/handoffs/tt-strategy-create-7-register.md

files_modified: [<list>]
strategy_name: <name>
documentation_updated: true
status: complete

Outputs

• •Updated strategy definition: src/strategy.ts
• •Updated research skill: lev-research/SKILL.md
• •Test verification artifacts
• •7 handoff artifacts in .lev/pm/handoffs/tt-strategy-create-*.md

Validation

• • Strategy compiles (pnpm build passes)
• • All existing tests pass (pnpm test)
• • New strategy executes via CLI
• • At least 1 adapter responds with sources
• • Strategy name doesn't collide with existing
• • lev-research skill updated with new strategy mapping
• • All 7 handoff files created

Usage

/workflow tt-strategy-create legal "Optimized for legal research: case law, regulations, compliance"
/workflow tt-strategy-create news-pulse "Real-time news monitoring with social signal" --adapters brave,grok,hackernews --mode parallel
/workflow tt-strategy-create competitive-intel "Company/product competitive analysis" --synthesizer oracle

Technique Map

• •Role definition - Clarifies operating scope and prevents ambiguous execution.
• •Context enrichment - Captures required inputs before actions.
• •Output structuring - Standardizes deliverables for consistent reuse.
• •Step-by-step workflow - Reduces errors by making execution order explicit.
• •Edge-case handling - Documents safe fallbacks when assumptions fail.

Technique Notes

These techniques improve reliability by making intent, inputs, outputs, and fallback paths explicit. Keep this section concise and additive so existing domain guidance remains primary.

Prompt Architect Overlay

Role Definition

You are the prompt-architect-enhanced specialist for workflow-tt-strategy-create, responsible for deterministic execution of this skill's guidance while preserving existing workflow and constraints.

Input Contract

• •Required: clear user intent and relevant context for this skill.
• •Preferred: repository/project constraints, existing artifacts, and success criteria.
• •If context is missing, ask focused questions before proceeding.

Output Contract

• •Provide structured, actionable outputs aligned to this skill's existing format.
• •Include assumptions and next steps when appropriate.
• •Preserve compatibility with existing sections and related skills.

Edge Cases & Fallbacks

• •If prerequisites are missing, provide a minimal safe path and request missing inputs.
• •If scope is ambiguous, narrow to the highest-confidence sub-task.
• •If a requested action conflicts with existing constraints, explain and offer compliant alternatives.