Datarim 1.11: General Cleanup Before Long Term Memory

Datarim 1.11.0 is a consolidation release. After weeks of rapid feature development — reflection integration, installer hardening, description optimization — the framework needed a clean baseline before the next major addition: Long Term Memory.

This release synchronizes every public artifact — the repository, documentation, and this website — with the current runtime state. Here is what changed.

Multi-Task Disambiguation

Datarim now supports multiple active tasks running in parallel. Previously, activeContext.md tracked a single "Current Task." When operators worked across several projects simultaneously, this led to confusion about which task a command targeted.

Version 1.11 introduces the Task Resolution Rule — a 5-step process that every pipeline command follows:

1. If the user provided a task ID argument — use it directly
2. Read the active tasks list from activeContext.md
3. If zero active tasks — stop and suggest /dr-init
4. If one active task — use it automatically (backward compatible)
5. If multiple — prompt the operator to choose

The single-task case works exactly as before. The new behavior only activates when you genuinely have parallel work.

Automated Runtime Sync

The operating model has always been runtime-first: changes happen in $HOME/.claude/, then get curated into the repository. Previously this was a manual cp process. Now curate-runtime.sh automates the mechanical part:

• --interactive mode: prompts per file (default)
• --auto mode: copies diffs and new files, skips deletes
• --dry-run mode: preview without changes
• Automatic patch-bump of VERSION, CLAUDE.md, and README.md

Combined with the existing check-drift.sh advisory tool, the full sync workflow is now: detect drift → curate with review → verify zero drift.

Description Optimization

Every skill, agent, and command in Datarim has a description in its YAML frontmatter. These descriptions drive auto-discovery — when Claude decides which skill to load. But long descriptions waste context tokens without improving triggering accuracy.

Version 1.11 enforces a 155-character maximum per description and achieved a 63% reduction in total description characters. Two overlapping skills (seo-launch and marketing) were merged into a single go-to-market skill.

What is Next: Long Term Memory

This cleanup release establishes a verified baseline for the next major feature: Long Term Memory integration. The LTM project (tracked as LTM-0001 through LTM-0003) will give Datarim persistent memory across sessions — remembering project context, operator preferences, and lessons learned beyond what fits in a single conversation window.

The research phase (LTM-0001) is complete. The benchmark phase (LTM-0002) is next. Having a clean, synchronized framework state means the LTM integration will start from a known-good foundation rather than chasing drift.

Full Change List

• Task Resolution Rule for multi-task disambiguation (TUNE-0018)
• curate-runtime.sh automated sync script (TUNE-0016)
• Description optimization — 63% reduction (TUNE-0017)
• seo-launch + marketing merged into go-to-market (TUNE-0017)
• PRD waiver check for Level 3-4 tasks (TUNE-0009)
• /dr-plan fixture capture requirement for CLI/API parsing
• 4 new skills: reflecting, infra-automation, infrastructure-debugging, go-to-market
• Full README rewrite (both project and repository)
• All datarim.club pages updated to v1.11.0

See the changelog for migration notes.