Prompts & Resources

Guided workflows and static reference guides available to your AI assistant

Prompts

Prompts are guided workflows that walk the host LLM through a multi-step process. The server exposes two.

`create-incentive`

A unified incentive-creation workflow. Structure:

Pre-flight checks — auth({ action: "status" }) → auth({ action: "login" }) if needed → list_projects + set_active_project (or create_project if none) → get_ai_context to ground suggestions in the project domain.
Choose action type — token (swap / bonding curve / hold), custom event, custom instruction, or Dune query. Each path has its own prerequisites (e.g. custom event → create_custom_event + attach_custom_event; IDL → create_idl preview + commit; Dune → register_dune_event_source + Dune-MCP validation + wait for first daily ingestion).
Generate query — generate_incentive_query with the appropriate source and source-specific params. Each action has a compatible set of measures (volume, count, duration, or balance); the prompt carries a matrix of which combinations are valid.
Preview (optional) — preview_incentive_query to validate results against real data.
Configure the incentive — create_recurring_incentive with type (leaderboard / rebate / raffle / direct), emission (tokens or SOL), schedule, and type-specific params.
Preview & confirm — call with confirmed: false first to surface the 5% protocol fee (added on top of totalFundAmount) and 7-day claim window, then confirmed: true.

The prompt includes the action-measure compatibility matrix, leaderboard payout-pattern examples, and an error-recovery section (auth expired mid-flow, preview returns no data, Dune source not yet ingested, incentive creation failure).

To use the prompt, ask your AI assistant:

"Use the create-incentive prompt to guide me through creating a new incentive."

`ask-torque`

Guided routing for the ask_torque escape-hatch tool. Walks the host LLM through session checks, considering native tools first, clarifying ambiguity, forwarding context, and calling the tool. See the Ask Torque guide for the full narrative.

To use the prompt:

"Use the ask-torque prompt to ask the Torque assistant: why did our swap volume dip on Tuesday?"

Resources

Resources are static reference guides the host LLM loads on demand.

URI	What it contains
`torque://capabilities`	Complete tool inventory grouped by bucket (Auth & Session, API Keys, Projects, Custom Events, IDL, Dune, Query Builder, Incentives, Results, Context, Ad-hoc Q&A), a 4-step quick-start flow with "wait for X" checkpoints, and an "I want to… / Do this" common-task table.
`torque://workflows`	Four canonical pipelines — Token, Custom Event, Custom Instruction, Dune Query — each with a one-liner canonical pipeline and step list. Sub-flows for reviewing existing incentives and viewing results/payouts/downloads. Terminology block (incentive type / action / measure / recurring incentive / epoch).
`torque://incentive-types`	The four incentive types with formula variables, functions, intent-to-formula examples (fixed rank prizes, tiered ranges, equal splits, diminishing returns, capped payouts, flat rewards), and the MCP-input-vs-persisted-name mapping.
`torque://building-landing-pages`	Domain model + epoch status lifecycle (`UPCOMING → EVALUATING → CLAIMING → COMPLETED / FAILED`), public REST endpoints for frontends, epoch-level data tables (distribution config, offer metadata, timing), claim flow, auth considerations, and a recommended frontend/backend architecture split.
`torque://ask-torque/capabilities`	Deep dive on the assistant behind `ask_torque` — what tools it can call (indexed DB queries, SQL validation, JS analysis, strategy suggestions, market data, Dune, project skills, file interpretation, download links), what is suppressed in MCP (charts + suggestion chips), and categorised example prompts.

These resources are available to the assistant automatically. You don't need to load them manually — the assistant reads them when relevant context is needed.