How yingsun1004/stackshift works
StackShift is an instruction-only learning Skill for experienced engineers changing technology stacks. A learner’s goal and existing experience shape a transfer map, gap/risk analysis, and feasibility-checked learning path; the selected mode then guides Learn → Apply → Verify and evidence-based readiness. The shared behavior is packaged for Codex and Claude Code, though the README says evaluation depth differs. The README documents the capabilities and stages below; sampled source was limited to evaluation/requalification helpers, not the Skill implementation itself.
Main entry points: skills/stackshift Learn Apply Verify · skills/stackshift Claude package · SKILL.md Behavior contract
Components 16
Runtime packages
- Codex package
.agents/skills/stackshift - Claude package
.claude/skills/stackshift - Behavior contract
.agents/skills/stackshift/SKILL.md
Transfer planning
- Prior experience
.agents/skills/stackshift - Transfer map
.agents/skills/stackshift - Gap and risk map
.agents/skills/stackshift - Goal feasibility
.agents/skills/stackshift - Learning path
.agents/skills/stackshift
Learning and evidence
- Learning modes
.agents/skills/stackshift - Repository mapping
.agents/skills/stackshift - Learn Apply Verify
.agents/skills/stackshift - Readiness evidence
.agents/skills/stackshift - Journey persistence
.agents/skills/stackshift
Other components
- Learner
- Codex
- Claude Code
Connections 19
- Codex → Codex package: invokes
- Claude Code → Claude package: invokes
- Codex package → Behavior contract: packages
- Claude package → Behavior contract: shares behavior
- Learner → Prior experience: provides
- Learner → Goal feasibility: states goal
- Prior experience → Transfer map: classifies
- Transfer map → Gap and risk map: identifies deltas
- Gap and risk map → Goal feasibility: informs
- Goal feasibility → Learning path: bounds
- Learning path → Learning modes: shapes
- Learning modes → Learn Apply Verify: selects
- Learner → Repository mapping: requests repo help
- Repository mapping → Learn Apply Verify: grounds learning
- Learn Apply Verify → Readiness evidence: produces observations
- Learner → Readiness evidence: demonstrates work
- Readiness evidence → Learn Apply Verify: guides next action
- Learner → Journey persistence: consents
- Journey persistence → Learn Apply Verify: resumes checkpoint