ln-230-story-prioritizer

Paths: File paths (references/, ../ln-*) are relative to this skill directory.

Story Prioritizer

Type: L3 Worker Category: 2XX Planning

Evaluate Stories using RICE scoring with market research. Generate consolidated prioritization table for Epic.

Purpose & Scope

• Prioritize Stories AFTER ln-220 creates them
• Triage all Stories cheaply before doing deep research
• Research market size and competition only where it changes prioritization confidence
• Calculate RICE score for each Story
• Generate prioritization table (P0/P1/P2/P3)
• Output: docs/market/[epic-slug]/prioritization.md

When to Use

Use this skill when:

• Stories created by ln-220, need business prioritization
• Planning sprint with limited capacity (which Stories first?)
• Stakeholder review requires data-driven priorities
• Evaluating feature ROI before implementation

Do NOT use when:

• Epic has no Stories yet (run ln-220 first)
• Stories are purely technical (infrastructure, refactoring)
• Prioritization already exists in docs/market/

Input Parameters

| Parameter | Required | Description | Default | |-----------|----------|-------------|---------| | epic | Yes | Epic ID or "Epic N" format | - | | stories | No | Specific Story IDs to prioritize | All in Epic | | depth | No | Research depth (quick/standard/deep) | "standard" |

depth options:

• quick - 2-3 min/Story, 1 WebSearch per type
• standard - 5-7 min/Story, 2-3 WebSearches per type
• deep - 8-10 min/Story, comprehensive research

Output Structure

docs/market/[epic-slug]/
└── prioritization.md    # Consolidated table + RICE details + sources

Runtime Contract

MANDATORY READ: Load references/planning_worker_runtime_contract.md, references/coordinator_summary_contract.md MANDATORY READ: Load references/researchgraph_mcp_usage.md when Stories cite H/G/run IDs or project researchgraph evidence can change priority confidence.

Runtime family: planning-worker-runtime

Identifier:

• epic-{epicId}

Phases:

1. PHASE_0_CONFIG
2. PHASE_1_DISCOVERY
3. PHASE_2_LOAD_STORY_METADATA
4. PHASE_3_ANALYZE_STORIES
5. PHASE_4_GENERATE_PRIORITIZATION
6. PHASE_5_WRITE_SUMMARY
7. PHASE_6_SELF_CHECK

Summary contract:

• summary_kind=story-prioritization-worker
• payload includes epic_id, depth, stories_analyzed, priority_distribution, top_story_ids, prioritization_path, warnings
• managed mode writes to caller-provided summaryArtifactPath
• default managed artifact path pattern: .hex-skills/runtime-artifacts/runs/{parent_run_id}/story-prioritization-worker/ln-230--{identifier}.json

Table columns (from user requirements):

| Priority | Customer Problem | Feature | Solution | Rationale | Impact | Market | Sources | Competition | |----------|------------------|---------|----------|-----------|--------|--------|---------|-------------| | P0 | User pain point | Story title | Technical approach | Why important | Business impact | $XB | [Link] | Blue 1-3 / Red 4-5 |

Inputs

| Input | Required | Source | Description | |-------|----------|--------|-------------| | epicId | Yes | args, kanban, user | Epic to process |

Resolution: Epic Resolution Chain. Status filter: Active (planned/started)

Tools Config

MANDATORY READ: Load references/environment_state_contract.md, references/storage_mode_detection.md, references/input_resolution_pattern.md

Extract: task_provider = Task Management → Provider

Research Tools

| Tool | Purpose | Example Query | |------|---------|---------------| | WebSearch | Market size, competitors | "[domain] market size {current_year}" | | mcp__Ref | Industry reports | "[domain] market analysis report" | | hex-research | Local hypothesis, goal, and benchmark evidence | find_hypotheses, inspect_goal, find_runs for explicit H/G/run context | | Task provider | Load Stories | IF linear: list_issues / ELSE: Glob story.md | | Glob | Check existing | "docs/market/[epic]/*" |

Workflow

Phase 1: Discovery (2 min)

Objective: Validate input and prepare context.

Process:

1. Resolve epicId: Run Epic Resolution Chain per guide.

1. Load Epic details:
2. IF task_provider == "linear": get_project(query=epicId)
3. ELSE IF task_provider == "github": gh issue view {epicId} -R {REPO} --json number,title,body
4. ELSE: Read("docs/tasks/epics/epic-{N}-*/epic.md")
5. Extract: Epic ID, title, description

1. Auto-discover configuration:
2. Read docs/tasks/kanban_board.md for Team ID
3. Slugify Epic title for output path

1. Check existing prioritization:

`` Glob: docs/market/[epic-slug]/prioritization.md ``

• If exists: Ask "Update existing or create new?"
• If new: Continue

1. Create output directory:

``bash mkdir -p docs/market/[epic-slug]/ ``

Output: Epic metadata, output path, existing check result

Phase 2: Load Stories Metadata (3 min)

Objective: Build Story queue with metadata only and prepare rough scoring inputs for all Stories.

Process:

1. Query Stories from Epic:

IF task_provider == "linear": `` list_issues(project=Epic.id, label="user-story") ` ELSE IF task_provider == "github": ` gh api /repos/{O}/{R}/issues/{epic_num}/sub_issues --jq '.[].number' → for each: gh issue view {num} -R {REPO} --json number,title,state,labels → filter: label "user-story" ` ELSE (file mode): ` Glob("docs/tasks/epics/epic-{N}-/stories//story.md") ``

1. Extract metadata only:
2. Story ID, title, status
3. minimal Epic context if available
4. DO NOT load full descriptions yet

1. Filter Stories:
2. Exclude: Done, Cancelled, Archived
3. Include: Backlog, Todo, In Progress

1. Build processing queue:
2. Order by: existing priority (if any), then by ID
3. Count: N Stories to process

Output: Story queue (ID + title + minimal context), ~50-80 tokens/Story

Phase 3: Two-Pass Story Analysis

Objective: Score all Stories cheaply first, then spend deep research only on candidates where it changes the decision.

Critical: Keep maximum context to one full Story at a time even during deep research.

If a researchgraph layout exists, run local graph checks only for Stories whose priority depends on hypothesis status, goal coverage, benchmark evidence, or implementation readiness. Local graph evidence can change RICE confidence and risk; it does not replace market-size or competitor research.

Pass A: Cheap Triage For All Stories

For each Story, load only enough detail to estimate:

• customer problem
• rough solution shape
• likely reach
• likely impact
• likely effort
• initial confidence tier

Step 3.1: Load Story Description

IF task_provider == "linear":

// configured tracker provider: getStory(id=storyId)

ELSE IF task_provider == "github":

gh issue view {storyId} -R {REPO} --json number,title,body,state,labels

ELSE (file mode):

Read("docs/tasks/epics/epic-{N}-*/stories/us{NNN}-*/story.md")

Extract from Story:

• Feature: Story title
• Customer Problem: From "So that [value]" + Context section
• Solution: From Technical Notes (implementation approach)
• Rationale: From AC + Success Criteria

Step 3.2: Build rough RICE estimate

Use Story + Epic context to assign:

• rough Reach
• rough Impact
• rough Effort
• initial Confidence

Mark one of:

• full_research_required
• rough_estimate_ok
• borderline_needs_review

Send to Pass B only if:

• candidate looks P0/P1 on rough score
• confidence is low
• Story is near a priority threshold
• Story has strategic or market-sensitive uncertainty

Pass B: Selective Deep Research

Only for Stories selected in Pass A, run full external research.

Step 3.3: Research Market Size

**WebSearch queries (based