fix(marketplace): sync manifest with actual skill/agent files

docs(think): add quality gates for SOW and Spec sections

refactor(reviewers): update reviewer files to enhance clarity and consistency

refactor(skills): prune 3 skills, clarify wrapper intent, restructure verify prose to tables

refactor(skills): split SKILL.md into procedure index + reference for 4 skills

feat(reviewers): add prompt-reviewer for LLM prompt quality assessment

fix(agents,docs): align confidence threshold to 0.70 and sync stale docs

feat(skills,templates): add session ID tracking to audit, research, and think outputs

feat(crosscheck,audit): uncommitted mode, calibration filters, reviewer sharpening

docs(adr): add confidence metadata and reassessment triggers to ADR templates

feat(spec): integrate EARS syntax into spec template and reviewer pipeline

fix(summarize): prevent hallucination from context and remove tropes source URL

feat(skills): add /crosscheck cross-model review skill

chore(rtk): remove rtk auto-rewrite hook and all references

feat(backlog): automate backlog lifecycle with launchagent and pending.md staging