From orbit
Mines local session transcripts, written plans, assistant claims, tool evidence, and codebase state for Orbit audits, retrospectives, lookbacks, review-window requests, gap analysis, validation, remediation planning, and questions like “what did we miss,” “what changed,” “check sessions,” “compare to the plan,” or “did it actually get done.”
How this skill is triggered — by the user, by Claude, or both
Slash command
/orbit:gap-analysisThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Use this skill when the user asks to audit a project, look back through sessions, identify missed work, compare intent to plans, validate assistant claims, or produce remediation recommendations.
Use this skill when the user asks to audit a project, look back through sessions, identify missed work, compare intent to plans, validate assistant claims, or produce remediation recommendations.
Plans are evidence, not the complete historical record. Assistant claims are claims, not proof. Validate against tool evidence and current repo state before marking work complete.
~/.claude/projects/**/*.jsonl when available.Compare:
user intent → plan evidence → assistant claims → tool evidence → codebase truth → gaps
Default review window is 3 days. Support explicit windows:
--days N
--since YYYY-MM-DD
--from YYYY-MM-DD --to YYYY-MM-DD
Separate:
Never treat a plan omission as proof that something was never requested. Mark missing evidence as uncertainty.
Use flat scripts when useful:
orbit-audit audit --days 3 --project . --validate-codebase --dashboard
orbit-audit compare-plan --plan PLAN.md --project . --validate-codebase
orbit-audit validate-claims --project . --evidence .claude/audits/latest/evidence.json
Creates, edits, and optimizes skills for Claude Code, including drafting, evaluating with test prompts, iterating on performance, and improving skill descriptions for better triggering accuracy.
npx claudepluginhub krzemienski/orbit --plugin orbit