prompt-context-enginelisted

# prompt-context-engine ## Mission Auto-gate #3 of the v2.6 four-gate chain. Classify, restructure, and inject every incoming user prompt into an optimized, executable task brief before Zeref begins work. Prevent wasted tokens, rework cycles, and misaligned execution caused by unstructured or under-specified prompts. Preserve every named tool, repo, constraint, and edge case from the original prompt (zero context loss — `_shared/rules.md#R6`). ## Model and Environment Guidance | Setting | Value | Reason | |---|---|---| | Default model | Sonnet 4.6 | Classification and restructuring are intelligence-sensitive but not flagship-cost justified | | Effort level | medium | Sufficient for prompt analysis and restructuring | | Token tier | S-M | Classification pass is short; brief output is compact | | Activate on | Every major task automatically | This is a system gate, not an optional feature | ## Use This Skill When - Any major task begins (auto-activation; user does not need to invoke this skill) - Prompt is fragmented, stream-of-consciousness, multi-intent, or abstract - Prompt is missing a clear deliverable, format, or constraint - Prompt contains named tools, repos, or systems that must be preserved through restructuring - Cost tier mismatch is detected (e.g., CRITICAL prompt with LOW-budget framing) ## Do Not Use This Skill When - Prompt is a simple factual question, single-file edit, or direct command - User explicitly says "execute verbatim" or "no restructure" - Pr