| name | qec-research |
| description | High-level orchestration skill for QEC research workflows. Initializes and updates project state, keeps a durable research log, routes work to future domain skills and MCP backends, and enforces explicit validation and provenance. |
| version | 0.1.0 |
| author | QEC Research Skills |
| license | MIT |
| tags | ["QEC","research","orchestration","state","provenance"] |
| dependencies | [] |
QEC Research
What This Skill Is For
Use this skill to coordinate a QEC research task at the project level:
- initialize or update project state,
- capture findings, blockers, and decisions,
- route work toward the right future backend or skill,
- keep benchmark and formalization work tied to durable artifacts.
When To Use It Versus Alternatives
Use this skill when the user asks for overall project progress, research planning, repo structure, state updates, or cross-cutting workflow decisions.
Do not use this skill when a narrower domain skill exists and the task is purely local to that workflow.
Required Artifacts And Assumptions
Required files:
research-state.yamlresearch-log.md
Assumptions:
- project memory must be updated when major decisions are made,
- claims about benchmarks or proofs require explicit validation state,
- future backend choices should be recorded, not implied.
Standard Workflow Checklist
- Read
research-state.yaml.
- Read the latest entries in
research-log.md.
- Identify the current phase, blockers, and next actions.
- Decide whether the request is architecture, literature, circuits, decoding, benchmarking, or formalization.
- Record any major decision that changes scope, architecture, or priorities.
- Update
research-state.yaml and research-log.md before closing substantive work.
Validation Loop
Before closing work:
- confirm the state file still matches the schema,
- confirm next actions are concrete,
- confirm any new repo conventions are written down,
- confirm open questions are still visible.
Common Issues And Fixes
State drift
Problem: the repo structure evolves, but the state file still describes an older plan.
Fix: update current_focus, phase, decisions, and next_actions together.
Skill sprawl
Problem: too many skills get proposed before the core contract exists.
Fix: record the proposed skill in the backlog, but do not add it until it supports a real demo or workflow.
Implicit backend assumptions
Problem: a plan assumes Stim, ldpc, Lean, or QUITS without stating why.
Fix: add the backend decision and rationale to research-state.yaml.
References
