feat(strategy): add minimal strategy layer on top of company brain

[divo12]

Summary

Adds a minimal strategy layer that decomposes decisions into goals grounded in the existing company brain (book context). This implements the smallest viable slice of the strategy system from the research requirements.

Changes

• Strategy module (src/strategy/)

  • types.ts: Decision and Goal domain types with status tracking
  • parser.ts: Decision seed parser with validation (500 char max)
  • decomposer.ts: Goal decomposition grounded in book context via FTS5 search
  • store.ts: File-based persistence under openwiki/.strategy/

• CLI commands

  • stratiki strategy seed <description> - Seed a new decision and generate grounded goals
  • stratiki strategy list - List all seeded decisions and their goals

• Tests

  • test/strategy/parser.test.ts - Decision validation tests
  • test/strategy/decomposer.test.ts - Goal decomposition and ranking tests with fixture wikis

How it works

1. User seeds a decision via CLI or (future) page
2. Decision is parsed and validated
3. System searches book context for relevant knowledge
4. Decision is decomposed into goals
5. Goals are ranked by how well they're grounded in existing company knowledge
6. Results stored in .strategy/ alongside the book

Example usage

# Seed a decision
stratiki strategy seed "Improve API authentication security"

# Output shows goals ranked by grounding quality:
# - [rank 110] Improve API authentication security
#   Grounded in: /architecture/api.md, /operations/security.md
# - [rank 105] Add monitoring for auth failures
#   Grounded in: /operations/monitoring.md

# List all decisions
stratiki strategy list

Testing

• All new tests pass (parser validation, goal decomposition, ranking)
• Existing test failures are unrelated (sqlite FTS5 module issues that existed before this PR)
• pnpm format and pnpm lint pass cleanly

Notes

• Reuses company mode home (~/.stratiki) and book context
• No scheduler, no agent runtime, no Horizon dependencies
• Goals ranked by grounding quality - better grounded = higher priority
• Failed/stale goal tracking via status field (future use)
• Thin contract for future execution layer

Related

Builds on company mode from PR #12 (merged). Does not reopen PR #11 or import from langchain-ai/openwiki.

  

---

Summary by cubic

Adds a minimal strategy layer that turns seeded decisions into goals grounded in the company book, exposing two new CLI commands (stratiki strategy seed|list). Previously there was no way to record decisions or track goals; now users can seed a decision and get ranked goals backed by existing wiki knowledge, persisted in the shared stratiki home under strategy/. The latest merge from main left an unresolved conflict in src/cli/cli.tsx, so the branch is not currently mergeable.

New Features

• New src/strategy/ module with decision parsing (validates non-empty, ≤500 chars), goal decomposition via book context search, and file-based persistence.
• New CLI commands: stratiki strategy seed <description> generates and scores goals by grounding quality, and stratiki strategy list shows all decisions and goal counts.
• Goal ranking and status fields form a thin contract for a future execution layer; no scheduler or agent runtime is included.

^{Written for commit 0ec92c5. Summary will update on new commits.}