8eebe65809
Pass 3 is the FINAL phase of the 3-pass research campaign: project the v2-deobfuscated outputs to C11 or Python code that conveys the subject video's content. Track scaffold: - spec.md: 14 sections, 11 videos, per-language default, 4 + 3 verification criteria - plan.md: 3 phases, 11 tasks, Tier 2 + 4 Tier 3 sub-agents - metadata.json: scope, per-language default, hardware target (up to ), risk register - state.toml: 3 phases, 11 tasks, verification flags - README.md: track index TIER2_STARTER.md (the dispatch prompt for Tier 2): - 15 sections, self-contained - The 4 PRIMARY inputs to read in order (v2 lexicon, C11 convention, Pass 1/2 content, manual_slop) - The 11 videos with per-language default (10 C11 + 2 Python + 1 synthesis) - The per-video deliverables (C11 .c/.h + 3 docs; Python .py + 3 docs) - The 4 + 3 verification criteria - The commit discipline (per-file atomic) - The 6 open questions answered - The 7 risks - The 4 Tier 3 sub-agent prompts (per cluster) Per-language default: C11 for math/algorithms oriented; Python for probability/information-theoretic; synthesis in Python. Tier 2 may override per video.