Strategic planning process skills (/plan-initiatives through /plan-stories)

[rucka]

Resolution

Delivered by PR #118 (merged). Board: Done.

feat: strategic planning process skills.

Story Statement

As a product manager or staff engineer doing strategic planning with AI assistance
I want to invoke /plan-initiatives, /map-subdomains, /map-contexts, /plan-epics, /plan-stories process skills, with how-to files 03-07 thinned to orchestrators
So that strategic planning phases from initiative definition through story breakdown are executable, consistent skills

Where: dataset/.skills/process/plan-initiatives/, dataset/.skills/process/map-subdomains/, dataset/.skills/process/map-contexts/, dataset/.skills/process/plan-epics/, dataset/.skills/process/plan-stories/, .pair/knowledge/how-to/03-*.md through 07-*.md

Epic Context

Parent Epic: Agent Skills in Knowledge Base #97
Status: Refined
Priority: P1 (Should-Have)

Acceptance Criteria

Functional Requirements

1. Given a project with a PRD and bootstrap complete
   When /plan-initiatives is invoked
   Then the skill follows the initiative creation process: PRD analysis → prioritization (P0/P1/P2) → initiative creation → documentation, composing /write-issue with $type: initiative and initiative-template.md for PM tool integration

2. Given /plan-initiatives is invoked and some initiatives already exist in the PM tool
   When the skill evaluates the current state
   Then it detects existing initiatives, skips them, and creates only missing ones (idempotent execution)

3. Given initiatives exist
   When /map-subdomains is invoked
   Then the skill follows subdomain definition: business capability analysis → DDD classification (core/supporting/generic) → subdomain specification, reading collaboration/templates/subdomain-template.md and producing files in adoption/product/subdomain/ following that template structure

4. Given /map-subdomains is invoked and some subdomain files already exist
   When the skill evaluates existing files
   Then it detects already-defined subdomains, skips them, and defines only missing ones. Offers to update existing subdomains if the developer requests it.

5. Given subdomains defined
   When /map-contexts is invoked
   Then the skill follows bounded context definition: constraint extraction → context catalog → integration pattern validation, reading collaboration/templates/bounded-context-template.md and producing files in adoption/tech/boundedcontext/ following that template structure

6. Given /map-contexts is invoked and some bounded context files already exist
   When the skill evaluates existing files
   Then it detects already-defined contexts, skips them, and defines only missing ones.

7. Given bounded contexts defined
   When /plan-epics is invoked
   Then the skill follows epic breakdown: initiative selection → epic analysis → epic documentation, composing /write-issue with $type: epic and epic-template.md for PM tool

8. Given /plan-epics is invoked and some epics already exist for the selected initiative
   When the skill evaluates existing epics in PM tool
   Then it detects existing epics, skips them, and creates only missing ones.

9. Given epics defined
   When /plan-stories is invoked
   Then the skill follows story breakdown: epic selection → story identification (vertical slicing, INVEST) → story documentation, composing /write-issue with $type: story and user-story-template.md for PM tool

10. Given /plan-stories is invoked and some stories already exist for the selected epic
    When the skill evaluates existing stories
    Then it detects existing stories, skips them, and creates only missing ones.

11. Given /write-issue is invoked with $type: epic and epic content
    When the skill executes
    Then it reads epic-template.md (collaboration/templates/epic-template.md), formats the issue body following that template structure, and creates (no $id) or updates ($id provided) the issue in the adopted PM tool

12. Given /write-issue is invoked with $type: initiative and initiative content
    When the skill executes
    Then it reads initiative-template.md (collaboration/templates/initiative-template.md), formats the issue body following that template structure, and creates (no $id) or updates ($id provided) the issue in the adopted PM tool

13. Given how-to files 03-07 are thinned
    When a non-skill assistant reads any of them
    Then each remains readable with phase flows, HALT conditions, priority frameworks, and selection criteria

14. Given pair's existing subdomain files in adoption/product/subdomain/
    When this story is implemented
    Then all existing subdomain files are updated to conform to the new subdomain-template.md structure

15. Given pair's existing bounded context files in adoption/tech/boundedcontext/
    When this story is implemented
    Then all existing bounded context files are updated to conform to the new bounded-context-template.md structure

Business Rules

• Idempotent execution (cross-cutting, applies to ALL skills): every step in a skill MUST check current state before acting, and only execute missing actions. Re-invoking a skill on a partially completed workflow resumes from where it left off — never re-does already completed work.
• /write-issue extension: this story extends /write-issue (from #101) with two new types:
  • $type: epic → reads collaboration/templates/epic-template.md
  • $type: initiative → reads collaboration/templates/initiative-template.md
  • Both templates already exist in the repo. /write-issue type validation is updated to accept story | task | epic | initiative.
  • All /write-issue rules apply: $id for create/update, PM tool agnostic (no fallback), template-driven formatting.
• Template-driven design artifacts: subdomains and bounded contexts are design/documentation artifacts (not PM tool issues), but they follow the same template-driven pattern:
  • /map-subdomains reads collaboration/templates/subdomain-template.md → produces files in adoption/product/subdomain/
  • /map-contexts reads collaboration/templates/bounded-context-template.md → produces files in adoption/tech/boundedcontext/
  • These skills write adoption files directly (NOT via /write-issue, since these are not PM tool issues).
  • Templates are created in this story, extracting the structure already defined in the how-to files and existing adoption files.
• Pair adoption alignment: existing subdomain and bounded context files in pair's own adoption (adoption/product/subdomain/*.md, adoption/tech/boundedcontext/*.md) MUST be updated to conform to the new templates. This ensures pair's own KB is consistent with the templates it provides.
• All 5 skills follow the established process skill pattern from Implement task with skills (/implement, /verify-quality, /record-decision) #99/Code review with skills (/review, /verify-done) #100/Sprint planning with skills (/refine-story, /plan-tasks, /write-issue) #101
• Idempotency detection per skill:
  • /plan-initiatives: checks PM tool for existing initiatives (via /write-issue duplicate detection)
  • /map-subdomains: checks adoption/product/subdomain/ for existing files
  • /map-contexts: checks adoption/tech/boundedcontext/ for existing files
  • /plan-epics: checks PM tool for existing epics under the selected initiative (via /write-issue)
  • /plan-stories: checks PM tool for existing stories under the selected epic (via /write-issue)
• How-to thinning ratios (from analysis):
  • how-to-03 (262 lines → ~80 lines): keeps priority framework, HALT gates; loses initiative template usage, tool procedures
  • how-to-04 (196 lines → ~80 lines): keeps HALT gates, DDD classification framework; loses business analysis procedures, specification format
  • how-to-05 (299 lines → ~105 lines): keeps HALT gates, catalog-before-details; loses constraint extraction, context mapping procedures
  • how-to-06 (215 lines → ~55 lines): keeps HALT gates, priority filtering, Epic 0 rule; loses epic template filling, linking procedures
  • how-to-07 (308 lines → ~100 lines): keeps HALT gates, INVEST framework, vertical slicing principles; loses story pattern identification, template filling
• /plan-initiatives, /plan-epics, /plan-stories compose /write-issue for PM tool integration (with template-driven formatting and duplicate detection)
• /map-subdomains and /map-contexts produce adoption files directly following their respective templates

Edge Cases and Error Handling

• Prerequisites missing (e.g. no PRD for initiatives, no subdomains for contexts): skill HALTs — inherited from how-to HALT conditions
• Partial completion: re-invocation detects existing output and creates only missing items
• Output file conflicts: /map-subdomains and /map-contexts detect existing files, offer update on request
• PM tool items out of sync: /plan-initiatives, /plan-epics, /plan-stories detect existing issues via /write-issue duplicate detection
• Epic/initiative template not found: /write-issue HALTs with error indicating which template is missing
• Subdomain/bounded-context template not found: /map-subdomains or /map-contexts HALTs with error indicating which template is missing
• Unsupported $type before this story: if /write-issue receives $type: epic or $type: initiative before this story is implemented, it HALTs with "unsupported type" (expected, resolved by this story)

Definition of Done Checklist

Development Completion

• 5 process SKILL.md files created with valid frontmatter and idempotent checks
• /write-issue SKILL.md updated to support epic and initiative types with template references
• collaboration/templates/subdomain-template.md created — extracted from how-to-04 and existing subdomain files
• collaboration/templates/bounded-context-template.md created — extracted from how-to-05 and existing bounded context files
• /map-subdomains reads subdomain-template.md to produce adoption files
• /map-contexts reads bounded-context-template.md to produce adoption files
• Pair's existing subdomain files (adoption/product/subdomain/*.md) updated to conform to subdomain-template.md
• Pair's existing bounded context files (adoption/tech/boundedcontext/*.md) updated to conform to bounded-context-template.md
• How-to files 03-07 thinned to orchestration only (~420 total lines)
• Every skill step follows check→skip→act→verify pattern (idempotency)
• /plan-initiatives composes /write-issue with $type: initiative + initiative-template.md
• /plan-epics composes /write-issue with $type: epic + epic-template.md
• /plan-stories composes /write-issue with $type: story + user-story-template.md
• All 5 thinned how-to files readable as standalone
• All 5 skills invocable in Claude Code
• Code reviewed and merged

Quality Assurance

• Each skill follows established process skill pattern (including idempotency)
• Re-invocation of each skill with partial output → creates only missing items
• HALT conditions from how-to files preserved
• /write-issue $type: epic → issue body follows epic-template.md
• /write-issue $type: initiative → issue body follows initiative-template.md
• /write-issue existing story and task types still work correctly (no regression)
• /map-subdomains produces files following subdomain-template.md
• /map-contexts produces files following bounded-context-template.md
• Pair's existing subdomain files conform to subdomain-template.md (no content lost)
• Pair's existing bounded context files conform to bounded-context-template.md (no content lost)
• Thinned how-to files reviewed for standalone readability

Story Sizing and Sprint Readiness

Refined Story Points

Final Story Points: L(5)
Confidence Level: High
Sizing Justification: 5 SKILL.md + 5 how-to thinnings + /write-issue extension (2 new types) + 2 new templates (subdomain, bounded-context) + pair adoption files update (9 files) + idempotency per skill. Volume is high but pattern is fully established. Templates are extracted from existing structures. Adoption file updates are restructuring, not rewriting.

Sprint Fit Assessment

Sprint Fit: Yes — single sprint
Development Estimate: 3-4 days (batch processing, pattern established)
Testing Estimate: 1 day (invocation + re-invocation + /write-issue extension + template validation + adoption alignment + standalone readability)

Dependencies and Coordination

Story Dependencies

Prerequisite Stories: #98 (skill infrastructure), #101 (/write-issue for PM integration — this story extends it with epic + initiative types)
Dependent Stories: None — these complete the process skill axis

Validation and Testing Strategy

Acceptance Testing

• Invoke each of the 5 skills in Claude Code → verify phase process
• Re-invoke each skill with partial output already existing → verify creates only missing items
• Invoke /plan-initiatives → verify composes /write-issue with $type: initiative, issue body follows initiative-template.md
• Invoke /plan-epics → verify composes /write-issue with $type: epic, issue body follows epic-template.md
• Invoke /plan-stories → verify composes /write-issue with $type: story, issue body follows user-story-template.md
• Invoke /write-issue with $type: epic, no $id → verify creates issue following epic template
• Invoke /write-issue with $type: initiative, $id: 10 → verify updates issue following initiative template
• Invoke /map-subdomains → verify produces files following subdomain-template.md
• Invoke /map-contexts → verify produces files following bounded-context-template.md
• Verify pair's subdomain files updated to match subdomain-template.md — all content preserved, structure standardized
• Verify pair's bounded context files updated to match bounded-context-template.md — all content preserved, structure standardized
• Read all 5 thinned how-to files → verify standalone readability
• Verify HALT conditions work (e.g. invoke /map-contexts without subdomains)

Notes

Open questions resolved:

• Thin all 5 at once: yes, batch processing (pattern established)
• Idempotency: each skill checks existing output before acting. /plan-initiatives checks PM tool, /map-subdomains checks adoption files, etc. All follow check→skip→act→verify pattern.
• /write-issue extension: this story adds epic and initiative to /write-issue's supported types. Templates already exist (epic-template.md, initiative-template.md). After this story, /write-issue supports: story | task | epic | initiative.
• Subdomain and bounded context templates: new templates created (subdomain-template.md, bounded-context-template.md) extracted from the structure already defined in how-to-04/05 and existing adoption files. /map-subdomains and /map-contexts read these templates directly (not via /write-issue, since these are design artifacts, not PM tool issues).
• Pair adoption alignment: existing subdomain (5 files) and bounded context (4 files) files in pair's own adoption are updated to conform to the new templates. Content is preserved, structure is standardized.

These 5 skills complete the process axis — all 11 how-to phases have corresponding skills.

Design reference: docs/design/skills-in-knowledge-base.md sections 1, 2

---

Refinement Date: 2026-02-08
Review and Approval: Approved (2026-02-09)

Technical Analysis

Implementation Approach

Technical Strategy: Write 5 process SKILL.md files + thin 5 how-to files + extend /write-issue with 2 new types + create 2 new templates (subdomain, bounded-context) + update pair's existing adoption files to match new templates. High volume but low complexity — pattern is fully established by Circle 1 (#98-#101). Each skill follows the same structure: check existing output → read context → read template → execute phases for missing items → produce output following template. YAML frontmatter description MUST be single-line (quoted string, no >- multi-line) for VS Code prompts-diagnostics-provider compatibility.

Key Files Modified/Created:

• dataset/.skills/process/plan-initiatives/SKILL.md (idempotent: checks existing initiatives)
• dataset/.skills/process/map-subdomains/SKILL.md (idempotent: checks existing subdomain files, template-driven)
• dataset/.skills/process/map-contexts/SKILL.md (idempotent: checks existing context files, template-driven)
• dataset/.skills/process/plan-epics/SKILL.md (idempotent: checks existing epics)
• dataset/.skills/process/plan-stories/SKILL.md (idempotent: checks existing stories)
• dataset/.skills/capability/write-issue/SKILL.md — updated to add epic and initiative types with their template references
• dataset/.pair/knowledge/guidelines/collaboration/templates/subdomain-template.md — new, extracted from how-to-04 Phase 3 structure and existing subdomain files
• dataset/.pair/knowledge/guidelines/collaboration/templates/bounded-context-template.md — new, extracted from how-to-05 Phase 3 structure and existing bounded context files
• dataset/.pair/adoption/product/subdomain/*.md — updated to conform to subdomain-template.md
• dataset/.pair/adoption/tech/boundedcontext/*.md — updated to conform to bounded-context-template.md
• 5 how-to files thinned (03 through 07)

/write-issue Extension:

• Add epic and initiative to supported $type values
• Add template mappings:
  • $type: epic → collaboration/templates/epic-template.md
  • $type: initiative → collaboration/templates/initiative-template.md
• All existing /write-issue behavior unchanged (PM tool agnostic, $id create/update, no fallback)
• Supported types after this story: story | task | epic | initiative

New Templates (extracted from existing patterns):

• subdomain-template.md: Title + Classification, Business Purpose, Key Capabilities, Strategic Importance, Complexity Assessment, Data Ownership, Dependencies, Team Recommendations, Implementation Priority
• bounded-context-template.md: Title + Type, Subdomains Covered, Business Scope and Purpose, Relationships Between Bounded Context and Sub Domains, Integration Patterns, Data Ownership, Team Alignment, Ubiquitous Language, Quality Attributes

Pair Adoption Update:

• Existing subdomain files (5 files): update to match subdomain-template.md — fill missing sections, standardize structure
• Existing bounded context files (4 files): update to match bounded-context-template.md — fill missing sections, standardize structure

Estimated total lines removed: ~860 lines across 5 how-to files (from ~1280 to ~420)

Technical Risks and Mitigation

Risk	Impact	Probability	Mitigation Strategy
5 how-to thinnings at once = large diff	Low	Medium	Pattern established, each thinning independent, review per file
/map-subdomains and /map-contexts DDD content complex	Medium	Low	Content already well-structured in how-to files
Duplicate detection across PM tools unreliable	Medium	Low	Delegates to /write-issue which already handles duplicate detection (#101)
/write-issue modification introduces regression	Medium	Low	Only adds new types, no change to existing story/task behavior
Existing adoption files may lose content during template alignment	Medium	Low	Preserve all existing content, only restructure to match template sections

Task Breakdown

• T-1: Create subdomain + bounded-context templates, conform existing files
• T-2: Update /write-issue for epic + initiative types
• T-3: Create /plan-initiatives SKILL.md
• T-4: Create /map-subdomains SKILL.md
• T-5: Create /map-contexts SKILL.md
• T-6: Create /plan-epics SKILL.md
• T-7: Create /plan-stories SKILL.md
• T-8: Thin how-to files 03-07

Dependency Graph

T-1 ──┬── T-4
      └── T-5
T-2 ──┬── T-3
      ├── T-6
      └── T-7
T-3 + T-4 + T-5 + T-6 + T-7 ── T-8

AC Coverage

AC	Brief Label	Tasks
AC-1	/plan-initiatives composes PM	T-3
AC-2	/plan-initiatives idempotent	T-3
AC-3	/map-subdomains reads template	T-4
AC-4	/map-subdomains idempotent	T-4
AC-5	/map-contexts reads template	T-5
AC-6	/map-contexts idempotent	T-5
AC-7	/plan-epics composes PM	T-6
AC-8	/plan-epics idempotent	T-6
AC-9	/plan-stories composes PM	T-7
AC-10	/plan-stories idempotent	T-7
AC-11	/write-issue handles epic	T-2
AC-12	/write-issue handles initiative	T-2
AC-13	thinned how-to 03-07 readable	T-8
AC-14	subdomain files conform template	T-1
AC-15	BC files conform template	T-1

---

T-1: Create subdomain + bounded-context templates, conform existing files

Priority: P0 | Estimated Hours: 3h | Bounded Context: KB Management

Summary: Create subdomain-template.md and bounded-context-template.md. Update existing adoption files in adoption/product/subdomain/ and adoption/tech/boundedcontext/ to conform to templates.

Type: Feature Implementation

Description: Two new template files enable /map-subdomains and /map-contexts to produce standardized output. Existing subdomain and BC files must be updated to match the template structure so new files follow the same pattern.

Acceptance Criteria:

• Primary deliverable: 2 template files + existing files conformed
• Quality standard: templates follow existing template conventions in guidelines/collaboration/templates/
• Integration requirement: used by /map-subdomains (T-4) and /map-contexts (T-5)
• Verification method: existing files parseable with template structure

Implementation Approach:

• Files to Modify/Create:
  • dataset/.pair/knowledge/guidelines/collaboration/templates/subdomain-template.md — new
  • dataset/.pair/knowledge/guidelines/collaboration/templates/bounded-context-template.md — new
  • dataset/.pair/adoption/product/subdomain/*.md — update to conform
  • dataset/.pair/adoption/tech/boundedcontext/*.md — update to conform

Dependencies:

• Tasks: None (independent)

Implementation Steps:

1. Analyze existing subdomain and BC files for common structure
2. Create subdomain-template.md with standardized sections
3. Create bounded-context-template.md with standardized sections
4. Update existing adoption files to conform to templates

Testing Strategy:

• Manual: verify template structure, verify existing files conform

---

T-2: Update /write-issue for epic + initiative types

Priority: P0 | Estimated Hours: 2h | Bounded Context: KB Management

Summary: Extend /write-issue to support $type: epic and $type: initiative in addition to existing story and task.

Type: Feature Implementation

Description: Update dataset/.skills/capability/write-issue/SKILL.md Step 1 validation and Step 3 template resolution to accept epic and initiative types. Create corresponding templates if needed, or reference existing PM tool implementation guide conventions.

Acceptance Criteria:

• Primary deliverable: /write-issue accepts $type: epic and $type: initiative
• Quality standard: follows existing type extension pattern (Step 1, Step 3)
• Integration requirement: composed by /plan-initiatives, /plan-epics
• Verification method: SKILL.md parseable, new types in validation + template resolution

Implementation Approach:

• Files to Modify/Create:
  • dataset/.skills/capability/write-issue/SKILL.md — update (add epic, initiative types)
  • dataset/.pair/knowledge/guidelines/collaboration/templates/epic-template.md — new (if needed)
  • dataset/.pair/knowledge/guidelines/collaboration/templates/initiative-template.md — new (if needed)

Dependencies:

• Tasks: None (independent)

Implementation Steps:

1. Add epic and initiative to Step 1 supported types
2. Add template resolution for both types in Step 3
3. Create epic-template.md and initiative-template.md if not existing
4. Update extensibility section to reflect new types

Testing Strategy:

• Manual: verify /write-issue accepts new types

---

T-3: Create /plan-initiatives SKILL.md

Priority: P0 | Estimated Hours: 3h | Bounded Context: KB Management

Summary: Write the /plan-initiatives process skill. Composes /write-issue for PM tool integration. Idempotent: detects existing initiatives.

Type: Feature Implementation

Description: Create dataset/.skills/process/plan-initiatives/SKILL.md. Extracts operational content from how-to-03. Guides strategic initiative creation from PRD goals. Composes /write-issue with $type: initiative. Idempotent: detects existing initiatives in PM tool, skips already-created ones.

Acceptance Criteria:

• Primary deliverable: dataset/.skills/process/plan-initiatives/SKILL.md with valid frontmatter
• Quality standard: check→skip→act→verify pattern, composition interface
• Integration requirement: composes /write-issue; composable standalone
• Verification method: algorithm covers AC-1 and AC-2

Implementation Approach:

• Files to Modify/Create:
  • dataset/.skills/process/plan-initiatives/SKILL.md — new

Dependencies:

• Tasks: T-2 (/write-issue must support initiative type)

Implementation Steps:

1. Extract operational content from how-to-03
2. Write algorithm with idempotent initiative detection
3. Add /write-issue composition for initiative creation
4. Add output format, graceful degradation

Testing Strategy:

• Manual: read SKILL.md, verify algorithm completeness

---

T-4: Create /map-subdomains SKILL.md

Priority: P0 | Estimated Hours: 3h | Bounded Context: KB Management

Summary: Write the /map-subdomains process skill. Reads subdomain template, produces subdomain files. Idempotent: detects existing subdomain definitions.

Type: Feature Implementation

Description: Create dataset/.skills/process/map-subdomains/SKILL.md. Extracts operational content from how-to-04. Guides domain decomposition producing subdomain catalog files in adoption/product/subdomain/. Uses subdomain-template.md from T-1. Idempotent: detects existing files, offers update.

Acceptance Criteria:

• Primary deliverable: dataset/.skills/process/map-subdomains/SKILL.md with valid frontmatter
• Quality standard: check→skip→act→verify, template-driven output
• Integration requirement: uses subdomain-template.md; composable by /bootstrap
• Verification method: algorithm covers AC-3 and AC-4

Implementation Approach:

• Files to Modify/Create:
  • dataset/.skills/process/map-subdomains/SKILL.md — new

Dependencies:

• Tasks: T-1 (subdomain-template.md must exist)

Implementation Steps:

1. Extract operational content from how-to-04
2. Write algorithm: detect existing → analyze PRD → identify subdomains → create files using template
3. Add idempotent detection of existing subdomain files
4. Add output format, graceful degradation

Testing Strategy:

• Manual: read SKILL.md, verify template usage, verify idempotency

---

T-5: Create /map-contexts SKILL.md

Priority: P0 | Estimated Hours: 3h | Bounded Context: KB Management

Summary: Write the /map-contexts process skill. Reads bounded-context template, produces BC files. Idempotent: detects existing BC definitions.

Type: Feature Implementation

Description: Create dataset/.skills/process/map-contexts/SKILL.md. Extracts operational content from how-to-05. Guides bounded context definition from subdomain catalog. Uses bounded-context-template.md from T-1. Idempotent: detects existing files, offers update.

Acceptance Criteria:

• Primary deliverable: dataset/.skills/process/map-contexts/SKILL.md with valid frontmatter
• Quality standard: check→skip→act→verify, template-driven output
• Integration requirement: uses bounded-context-template.md; composable by /bootstrap
• Verification method: algorithm covers AC-5 and AC-6

Implementation Approach:

• Files to Modify/Create:
  • dataset/.skills/process/map-contexts/SKILL.md — new

Dependencies:

• Tasks: T-1 (bounded-context-template.md must exist)

Implementation Steps:

1. Extract operational content from how-to-05
2. Write algorithm: detect existing → read subdomains → map to contexts → create files using template
3. Add idempotent detection of existing BC files
4. Add output format, graceful degradation

Testing Strategy:

• Manual: read SKILL.md, verify template usage, verify idempotency

---

T-6: Create /plan-epics SKILL.md

Priority: P0 | Estimated Hours: 3h | Bounded Context: KB Management

Summary: Write the /plan-epics process skill. Composes /write-issue for PM tool integration. Idempotent: detects existing epics.

Type: Feature Implementation

Description: Create dataset/.skills/process/plan-epics/SKILL.md. Extracts operational content from how-to-06. Guides epic breakdown from initiatives. Composes /write-issue with $type: epic. Idempotent: detects existing epics, skips already-created ones.

Acceptance Criteria:

• Primary deliverable: dataset/.skills/process/plan-epics/SKILL.md with valid frontmatter
• Quality standard: check→skip→act→verify, composition interface
• Integration requirement: composes /write-issue; composable standalone
• Verification method: algorithm covers AC-7 and AC-8

Implementation Approach:

• Files to Modify/Create:
  • dataset/.skills/process/plan-epics/SKILL.md — new

Dependencies:

• Tasks: T-2 (/write-issue must support epic type)

Implementation Steps:

1. Extract operational content from how-to-06
2. Write algorithm with idempotent epic detection
3. Add /write-issue composition for epic creation
4. Add output format, graceful degradation

Testing Strategy:

• Manual: read SKILL.md, verify algorithm completeness

---

T-7: Create /plan-stories SKILL.md

Priority: P0 | Estimated Hours: 3h | Bounded Context: KB Management

Summary: Write the /plan-stories process skill. Composes /write-issue for PM tool integration. Idempotent: detects existing stories.

Type: Feature Implementation

Description: Create dataset/.skills/process/plan-stories/SKILL.md. Extracts operational content from how-to-07. Guides story breakdown from epics with INVEST validation. Composes /write-issue with $type: story. Idempotent: detects existing stories, skips already-created ones.

Acceptance Criteria:

• Primary deliverable: dataset/.skills/process/plan-stories/SKILL.md with valid frontmatter
• Quality standard: check→skip→act→verify, INVEST validation, composition interface
• Integration requirement: composes /write-issue; composable standalone
• Verification method: algorithm covers AC-9 and AC-10

Implementation Approach:

• Files to Modify/Create:
  • dataset/.skills/process/plan-stories/SKILL.md — new

Dependencies:

• Tasks: T-2 (/write-issue must support story type — already does)

Implementation Steps:

1. Extract operational content from how-to-07
2. Write algorithm with INVEST validation and idempotent story detection
3. Add /write-issue composition for story creation
4. Add output format, graceful degradation

Testing Strategy:

• Manual: read SKILL.md, verify INVEST validation, verify idempotency

---

T-8: Thin how-to files 03-07

Priority: P0 | Estimated Hours: 4h | Bounded Context: KB Management

Summary: Thin 5 how-to files (03 through 07) removing operational content moved to skills T-3 through T-7. Keep orchestration, HALT gates, references. ~860 lines removed total.

Type: Refactoring

Description: Thin each how-to file to orchestration-only: phase flow, HALT conditions, approval gates, references, next steps. Operational content (procedures, templates, checklists) moved to corresponding SKILL.md files. Each thinned file must remain standalone readable.

Acceptance Criteria:

• Primary deliverable: 5 how-to files thinned (~65% reduction each)
• Quality standard: standalone readable, phase flow preserved, HALT conditions intact
• Integration requirement: references corresponding skill for operational details
• Verification method: read each file, verify orchestration completeness

Implementation Approach:

• Files to Modify:
  • dataset/.pair/knowledge/how-to/03-how-to-create-and-prioritize-initiatives.md
  • dataset/.pair/knowledge/how-to/04-how-to-define-subdomains.md
  • dataset/.pair/knowledge/how-to/05-how-to-define-bounded-contexts.md
  • dataset/.pair/knowledge/how-to/06-how-to-breakdown-epics.md
  • dataset/.pair/knowledge/how-to/07-how-to-breakdown-user-stories.md

Dependencies:

• Tasks: T-3, T-4, T-5, T-6, T-7 (content must exist in skills first)

Implementation Steps:

1. For each how-to, identify operational content now in corresponding skill
2. Remove operational sections, replace with brief phase summaries + skill reference
3. Preserve orchestration: overview, session state, HALT gates, references, next steps
4. Verify standalone readability for each file

Testing Strategy:

• Manual: read all 5 thinned files end-to-end