Spec to Tickets

What it does

Spec to Tickets turns a caller-provided PRD/spec into a structured DRAFT proposal of tickets — a parent epic plus issues with titles, descriptions, acceptance criteria, estimates, labels, priorities, and dependency hints. It is connector-free, draft-only, and approval-gated: the output is always a proposal with created_ids=[] and requires_confirmation=true, and the agent NEVER writes to any issue tracker. The PRD is treated strictly as untrusted data. It is deterministically parsed into sections and grounded ticket skeletons (with priority and dependency hints); an LLM then drafts the full ticket content, with a deterministic template fallback when no model is available. Tickets are grounded in the PRD — no invented requirements — and any instructions embedded in the PRD are neutralized and ignored. Tools: - run_full — end-to-end: analyze the PRD into grounded skeletons, draft tickets via the LLM (template fallback), and return a DRAFT-ONLY TicketProposal. Never creates anything. - analyze_prd — pure deterministic analysis (no LLM): parse the PRD into sections and grounded ticket skeletons with priority + dependency hints. - create_tickets — approval-gate stub: always refuses and returns the draft proposal plus a confirmation-required notice. Never writes. - list_capabilities — static capabilities: tools, archetype, and guardrail invariants. - plan_inputs — interview helper returning questions, JSON schema, and a ready-to-edit example for a tool. Provide prd_text (the spec); optionally a team label and create_epic flag. Array/object arguments may be passed as JSON or a JSON string. Missing required inputs return a structured needs_input payload (questions + schema + example) instead of erroring, and upgrade to a native elicitation form on clients that support it.

Example prompts

• Turn this PRD into a draft epic + tickets with acceptance criteria, estimates, and dependency hints.
• Run analyze_prd on this spec — just the sections and grounded ticket skeletons with priorities, no LLM.
• Break this feature spec into issues for the Platform Team and flag which tickets depend on which.
• Call create_tickets on this proposal — confirm it refuses and returns the confirmation-required draft.
• Use plan_inputs for run_full and show me the questions, schema, and a ready-to-edit example.

Tools (5)

Tools the agent exposes — your AI client calls them automatically when it needs them.

• run_full — Run end-to-end: take a caller-provided PRD/spec text, analyze it into grounded ticket skeletons, draft tickets via the LLM (with a deterministic template fallback), and return a DRAFT-ONLY TicketProposal (created_ids=[], requires_confirmation=true). Never creates anything.
• analyze_prd — Pure deterministic analysis (no LLM): parse the PRD into sections and grounded ticket skeletons with priority + dependency hints.
• create_tickets — STUB / APPROVAL GATE: would create the proposed tickets in an issue tracker, but this connector-free build NEVER writes. Always refuses and returns the draft proposal plus a confirmation-required notice.
• list_capabilities — List the agent's static capabilities: tools, archetype, and guardrail invariants.
• plan_inputs — Plan/brainstorm the inputs for a tool: returns the questions, schema and a ready-to-edit example.