Skip to content

TODOS.md

Latest commit

 

History

History
82 lines (72 loc) · 4.8 KB

TODOS.md

File metadata and controls

82 lines (72 loc) · 4.8 KB
Raw

TODOS

v1.1: Manifest Versioning SHIPPED (v1.2.0)

Shipped in v1.2.0. bin/idstack-migrate handles schema migrations. All 7 skill preambles call it automatically.

v2: Cross-Domain Quality Checks SHIPPED (v1.2.0)

Shipped in v1.2.0. Four checks added to /course-quality-review: cognitive load, multimedia principles, feedback quality, expertise reversal. Plus: fix-link integration, quick-win prioritization, score trending, shareable quality report, and per-category scoring breakdown.

v2: Template System

At 5+ skills, evaluate SKILL.md.tmpl -> SKILL.md template system for shared preamble (manifest read/write, cross-cutting concerns, evidence tier definitions). Why: Duplication across 3 skills is acceptable. Across 8+ it becomes a maintenance burden. Depends on: Decision to build additional skills beyond v1's 3.

v2: Additional Skills (from literature synthesis)

The remaining skill suite from the evidence synthesis includes 4 more skills:

  • id-model-selector (Domain 1) — context-driven model/framework recommendation
  • content-sequencing (Domain 4) — cognitive load-aware sequencing and chunking
  • media-selection (Domain 6) — multimedia principle application and violation flagging
  • evaluation-design (Domain 8) — Kirkpatrick+ evaluation planning Why: Extends coverage for specialized ID workflows. Each backed by evidence synthesis. Depends on: User feedback on which capabilities are most requested. Note: assessment-design, course-builder, and course-export are now shipped (v1.1). learner-analysis is partially covered by /needs-analysis (Level 3) and /course-import. prototype-iteration overlaps with the build/export pipeline.

v2: LMS-Specific API Integrations

Add Blackboard (xAPI), Moodle (web services), and D2L (Valence API) as import sources alongside Canvas. The IMS Common Cartridge format covers these generically, but LMS-specific APIs unlock richer data (rubrics, analytics, student engagement). Why: Each LMS has unique data that cartridge export doesn't capture. Depends on: /course-import Canvas API path stable and tested. User demand. Priority: P3

v2: SCORM/xAPI Package Import SHIPPED (v1.4.0)

Shipped in v1.4.0. SCORM 1.2/2004 import added as Path E in /course-import. SCORM 1.2 export added as Path C in /course-export. PDF import also added (Path D).

v2: Standardize Next-Step Formatting

Normalize the "next step" section format across all SKILL.md files. Currently inconsistent: "Next step:" (singular), "Next steps:" (plural), "Recommended next steps:" (course-import). Why: Consistency improves the user experience and makes the pipeline feel cohesive. Depends on: Nothing. Low-effort cleanup. Priority: P3

v2: Multi-Platform Install (Gemini CLI + Codex CLI)

Add install support for Gemini CLI (/.gemini/skills/) and Codex CLI (/.agents/skills/). All three platforms use the same SKILL.md format (name + description frontmatter). Key challenges: hardcoded ~/.claude/ paths in all 7 SKILL.md preambles need refactoring to platform-agnostic paths (IDSTACK_HOME env var or relative resolution), allowed-tools frontmatter uses Claude Code tool names, and tool name compatibility needs verification. Why: Expands addressable market to Gemini CLI and Codex CLI users. Depends on: User feedback confirming demand (via Google Form). Spike test on one SKILL.md to verify compatibility before full implementation. Priority: P2

v2: Sub-Agent Architecture for Context Efficiency

Skills like /red-team run 5 adversarial dimensions sequentially in the main context. Each dimension could dispatch as a sub-agent, keeping the main session lean. Carl Vellotti's research shows sub-agents can reduce main session context from 25% to 16.5% for equivalent work. For /red-team: dimensions 1-5 as parallel sub-agents, results aggregated in the main session for scoring. Same pattern applies to /accessibility-review (WCAG + UDL could be parallel sub-agents) and /course-quality-review (QM + CoI). Why: Context efficiency directly affects session length. A session that compacts after 5 messages vs one that lasts 30 is a completely different user experience. Depends on: Understanding how each AI CLI handles sub-agent dispatching (Claude Code Agent tool, Gemini CLI equivalent, Codex subagents). Template system (v2) would make this easier to implement consistently across skills. Priority: P2

v3: Bidirectional LMS Sync

Push changes back to Canvas (and eventually other LMS) via API. After /course-quality-review identifies issues and /learning-objectives generates better ILOs, push the improvements back to the LMS. The output IS the course. Why: The 10x vision. Eliminates the translation step from design doc to LMS. Depends on: /course-import stable, Canvas API write operations investigated, conflict handling designed, institutional partnerships.