Releases: savvides/idstack
Releases · savvides/idstack
v3.2.0.0
Full Changelog: v3.1.0.0...v3.2.0.0
v3.1.0.0 — DESIGN.md and design-system reconciliation
The visual system that ships in the report stylesheet and the landing page is now documented at the repo root in DESIGN.md. Skills, contributors, and reviewers read it before touching anything visual. CLAUDE.md and AGENTS.md route to it.
Added — DESIGN.md and design-system reconciliation
DESIGN.mdat the repo root. Source of truth for fonts, colors, spacing, border radii, motion timing, and aesthetic direction. North-star is Proof — the artifact should read like a peer-reviewed clinical reference. Includes anti-patterns to never ship and a dated decisions log.- Web fonts swapped to publication-grade. Source Serif 4 (display + body), Public Sans (UI/labels/badges), JetBrains Mono (citations/IDs/code), loaded via Google Fonts
<link>indocs/index.html,templates/report.html.tmpl, andtemplates/index.html.tmpl. System-font fallbacks preserved so pages stay legible offline. - Palette shift: ivory replaces parchment. Background goes from
#fbfaf6(parchment-warm) to#faf8f3(pristine ivory). Reads as "good paper," not "old book." - Second annotation color added. Prussian blue
#1d4a5ejoins library-stamp red#7a1f1f. Links and citation cross-references now use blue; the kicker / primary marks stay red. Mirrors the two-pen academic-editor convention. - Sharper corners across the surface. Finding cards and other card surfaces are now
border-radius: 0. Buttons, badges, chips, install snippets normalize to2px. The publication signal at the expense of consumer-SaaS softness. - Tier-badge weights amplified. T1 / T2 render bold (the strongest evidence reads heaviest); T5 stays light grey (expert opinion reads weakest). Hex values unchanged.
Changed
templates/assets/idstack.css— reconciled toDESIGN.mdtokens (ivory bg, prussian-blue links, sharp finding cards, new font tokens). Shipped to every course folder as before.docs/index.html— same token swap; UI chrome (eyebrows, badges, copy button, footer form, brand-beta, severity chips, pipeline labels) now uses Public Sans via--font-ui.templates/report.html.tmpl/templates/index.html.tmpl—<head>now includes Google Fonts preconnect + stylesheet link for Source Serif 4 / Public Sans / JetBrains Mono.CLAUDE.md/AGENTS.md(viatemplates/agent-context.md) — new Design system section instructing skills to readDESIGN.mdbefore any UI decision.
Notes
- No path, manifest, or API changes; minor version bump because the design system is now load-bearing for new skills and reviewers.
- Reports written by older versions still render correctly with the new stylesheet — the existing class hooks (
finding sev-{severity},tier-badge tier-T{N},sev-badge,citation) are unchanged.
Full Changelog: v2.0.0...v3.1.0.0
v2.0.0 — Pipeline Orchestrator, Cross-Course Intelligence, Sub-Agents
One command runs the full instructional design pipeline. Learnings transfer between courses. Review skills run in parallel.
What's new
Pipeline Orchestrator
/idstack pipelinechains all 8 design skills automatically. Auto-skips completed skills, shows pipeline status, pause and resume anytime.
Namespace
- All skills now invoked via
/idstack <skill>(e.g.,/idstack needs-analysis). No more name collisions with other skill packages. - Setup creates a single symlink. Cleans up legacy individual links automatically.
Cross-Course Intelligence
- Learnings from one course appear in another via
~/.idstack/global/learnings.jsonl. - Keyword search with
--keywordand--cross-projectflags. /idstack learnskill: search, list, delete, promote, and export learnings.
Designer Profile & Preferences
- Create
~/.idstack/profile.yamlwithexperience_level: novice|intermediate|expert. Skills adapt explanation depth. - Manifest preferences (schema v1.3): verbosity, export format, preferred LMS.
Sub-Agent Architecture (Claude Code only)
/idstack red-team: 5 parallel dimension agents/idstack accessibility-review: 2 parallel agents (WCAG + UDL)/idstack course-quality-review: 3 parallel agents (QM + CoI + alignment)/idstack course-builder: adversarial spec review loop post-generation- Graceful sequential fallback on other platforms.
Course Readiness Dashboard
bin/idstack-status --readinessshows pre-export gate: quality score, red-team findings, accessibility score, with pass/fail verdict.- Integrated into
/idstack course-export.
Infrastructure
IDSTACK_HOMEenv var. All paths portable. Foundation for multi-platform support.- Chained schema migration v1.0/v1.1/v1.2 → v1.3 in one pass.
- New bin scripts:
idstack-learnings-delete,idstack-learnings-promote.
Breaking changes
- Skills are now invoked as
/idstack <skill>instead of/<skill>. Aftergit pull, re-run./setupto clean up old symlinks.
Install / Update
New installs:
git clone https://github.com/savvides/idstack.git ~/.claude/skills/idstack && cd ~/.claude/skills/idstack && ./setup
Existing users:
cd ~/.claude/skills/idstack && git pull && ./setup
The ./setup step is required — it cleans up legacy individual symlinks and creates the new single /idstack entry point.
Full changelog: CHANGELOG.md
v1.5.0 — Course Memory
idstack now remembers your design sessions.
What's new
- Course memory. Each skill logs what it did to
.idstack/timeline.jsonl. The next session starts with a welcome-back message: quality score trend, last skill run, suggested next step. Your course context compounds across sessions. - Learnings. When a skill discovers something specific to your project (Canvas uses custom rubric formatting, SCORM packages from Rise need special handling), it stores the discovery in
.idstack/learnings.jsonl. Future sessions surface relevant learnings automatically. - Pipeline guidance. Skills suggest the next step based on what you've already completed, replacing static "Next step:" text with data-driven recommendations.
bin/idstack-statusprints a course health dashboard: skills completed, quality score trend, dimension breakdown, and suggested next skill.- Template system. Shared preamble code maintained in one place (
templates/preamble.md). Contributors edit.tmplfiles and runbin/idstack-gen-skillsto regenerate. - python3 recommended. Course memory features work best with python3 (score trends, search filtering, JSON validation). Without it, basic timeline logging still works via bash fallback.
Install / Update
If you already have idstack, it auto-updates on next skill invocation. For new installs:
git clone https://github.com/savvides/idstack.git ~/.claude/skills/idstack && cd ~/.claude/skills/idstack && ./setup
Full changelog: CHANGELOG.md
v1.4.0
Full Changelog: v1.3.0...v1.4.0
v1.3.0
Full Changelog: v1.2.1...v1.3.0
v1.2.1
Full Changelog: v1.2.0...v1.2.1
v1.2.0
Full Changelog: v1.1.1...v1.2.0
v1.1.1
Fixed
- Pipeline ordering: quality review now comes before export in all diagrams and skill next-step recommendations
- learning-objectives next step now correctly points to assessment-design
- course-builder next steps reordered: quality review first, export second
- course-quality-review now recommends course-export as the next step when audit passes
Added
- VERSION file for formal semver tracking
- TODOS.md: tracking next-step formatting standardization (P3)
v1.1.0 — Full Lifecycle: Assess, Build, Export
idstack v1.1.0
Three new skills complete the full instructional design lifecycle.
New Skills
/assessment-design— Evidence-based rubrics, feedback strategies, and formative checkpoints aligned to Bloom's taxonomy. Applies Nicol's 7 principles of good feedback practice./course-builder— Generates complete course content (syllabus, modules, assignments, rubrics) from the manifest. Content follows cognitive load principles and adapts to learner expertise./course-export— Exports to any LMS via IMS Common Cartridge (.imscc) or pushes directly to Canvas via REST API. The output IS the course.
Full Pipeline
/needs-analysis → /learning-objectives → /assessment-design → /course-builder → /course-export
↑ │
└────────── /course-import ──────────── /course-quality-review ──────────────┘
Seven skills. Import from any LMS, design with evidence, build content, export back to any LMS.
Upgrade
cd path/to/idstack && git pull && ./setup