prioritize-epics

name	prioritize-epics
description	Use when the user wants to (re)prioritize an existing backlog of epics — assigns P0–P4 priorities across all open epics based on a stakeholder-defined North Star, risk tolerance, tradeoffs, and out-of-scope criteria. Runs a 5-phase pipeline that (1) assesses each epic's clarity, distills the overall vision, compares it to current state, and maps risk hotspots; (2) interviews the stakeholder via Socratic dialogue to capture the prioritization rubric; (3) coarse-sorts epics into North Star / optional additions / out of scope using batched sub-agent evaluation; (4) granular-sorts within each bucket into P0–P4; (5) offers to brainstorm any P0/P1 epics that lack the `brainstorm:complete` tag. Trigger phrases include 'prioritize the epics', 'reprioritize the backlog', 'set epic priorities', 'sort the backlog', 'rank the epics', 'what should we work on next at the epic level'.
user-invocable	true
allowed-tools	Read, Write, Edit, Glob, Grep, Bash

Requires Agent tool. If running as a sub-agent (Agent tool unavailable), STOP and return: "ERROR: /dso:prioritize-epics requires Agent tool; invoke from orchestrator."

Prioritize Epics

Act as a Senior Product Manager. Assign P0–P4 priorities across an existing backlog of open epics using a stakeholder-defined rubric. This skill prioritizes existing epics; it does not create new ones. Use /dso:roadmap or /dso:brainstorm to create epics.

Usage

/dso:prioritize-epics

Interactive. Walks through 5 phases with explicit user confirmation between each. Supports dryrun mode via /dso:dryrun /dso:prioritize-epics.

Scope

• In scope: assigning priorities (P0–P4) to existing open epics; adding clarifying comments to epics that lacked enough information; offering to brainstorm any P0/P1 epic that is not yet brainstormed.
• Out of scope: creating new epics, modifying epic descriptions or success criteria (beyond appending a Phase 1 clarification comment), decomposing epics into stories/tasks.

Layout

This skill's logic is split across phase files. Load each on demand:

File	When to read
phases/phase-1-assessment.md	Phase 1 — project assessment
phases/phase-2-stakeholder.md	Phase 2 — stakeholder interview
phases/phase-3-coarse-sort.md	Phase 3 — coarse sorting
phases/phase-4-granular-sort.md	Phase 4 — granular sorting
phases/phase-5-brainstorming.md	Phase 5 — brainstorming follow-up
prompts/epic-clarity-review.md	Phase 1 Step 1 sub-agent prompt
prompts/vision-summary.md	Phase 1 Step 3 sub-agent prompt
prompts/vision-gap.md	Phase 1 Step 4 sub-agent prompt
prompts/risk-hotspots.md	Phase 1 Step 5 sub-agent prompt
prompts/coarse-sort-batch.md	Phase 3 Step 1 sub-agent prompt

Session State

Maintain these variables across phases:

Variable	Set in	Used by
OUT_OF_SCOPE_IDS	Phase 1 Step 2 (epics the user said are out of scope)	Phase 3 (excluded from batch sort)
CLARIFIED_IDS	Phase 1 Step 2 (epics that got a clarification comment)	Logging only
VISION_SUMMARY	Phase 1 Step 3	Phase 2 framing, Phase 3 subagent prompt
STATE_VS_VISION_GAP	Phase 1 Step 4	Phase 2 framing
RISK_HOTSPOTS	Phase 1 Step 5	Phase 2 framing, risk-tolerance discussion
NORTH_STAR	Phase 2 Step 2 (approved)	Phase 3, 4
RISK_TOLERANCE	Phase 2 Step 2 (approved)	Phase 3, 4
TRADEOFFS	Phase 2 Step 2 (approved)	Phase 4
OUT_OF_SCOPE_DEFINITION	Phase 2 Step 2 (approved)	Phase 3, 4
AVOID_RISK	Phase 2 Step 2 (boolean derived from RISK_TOLERANCE)	Phase 3 subagent prompt, Phase 3 Step 2
BUCKETS	Phase 3 (after dependency adjustments)	Phase 4

Phase Gate Pattern

After each phase, emit:

=== Phase [N] Complete ===

[Summary of what was accomplished in this phase]

PHASE GATE QUESTION:
[Specific confirmation question from the phase]

Do NOT proceed until user responds.

Execution Framework

Read each phase file in order. Do NOT skip phases. Do NOT proceed to the next phase without the user clearing the phase gate.

1. Read phases/phase-1-assessment.md and execute it.
2. Read phases/phase-2-stakeholder.md and execute it.
3. Read phases/phase-3-coarse-sort.md and execute it.
4. Read phases/phase-4-granular-sort.md and execute it.
5. Read phases/phase-5-brainstorming.md and execute it.

Guardrails

• One question at a time. All Socratic dialogue loops (Phase 1 Step 2, Phase 2 Step 1) ask exactly one question per turn. Never batch multiple questions.
• Never re-ask answered questions. Before each question, review prior turns for semantic duplicates.
• Never assign a priority without an approved rubric. Phase 2 Step 3 (approval gate) must clear before any priority is written.
• Never modify epic priorities mid-process. Priorities are written only in Phase 4 — except the Phase 1 Step 2 P4 assignment for user-declared out-of-scope epics, which is the explicit exception.
• Never write priorities to closed or archived epics. Filter all epic lists by --status=open,in_progress.
• Never create or decompose epics. This skill only re-prioritizes. If the user asks to create or decompose, point them at /dso:roadmap, /dso:brainstorm, or /dso:preplanning.
• Always use .claude/scripts/dso ticket edit <id> --priority=<n> to set priorities. Priorities are integers 0–4.

Ticket CLI Quick Reference

Action	Command
List open epics	.claude/scripts/dso ticket list --type=epic --status=open,in_progress --format=llm
Show one epic	.claude/scripts/dso ticket show <id>
Add a comment	.claude/scripts/dso ticket comment <id> "<body>"
Set priority	.claude/scripts/dso ticket edit <id> --priority=<0-4>
Dependency graph	.claude/scripts/dso ticket deps <id>
Check tag	inspect tags field in .claude/scripts/dso ticket show <id>

Always verify exact syntax via ${CLAUDE_PLUGIN_ROOT}/docs/ticket-cli-reference.md before invoking ticket commands.