modus-sync-knowledge
This is a silent executor for code tasks with confirmed long-term knowledge impact, not a user workflow. After using-modus initially determines positive impact, a safety review is still performed: the scope is determined by the actual diff of the current task, the responsibility by the body content, and the permission by human ownership; no initialization bundle is built or read.
Only attribute the actual changes made in this round; do not take over existing dirty changes in the workspace. If evidence is insufficient, silently analyze checkpoints without running fixed commands or outputting noop.
- Review the actual modifications in this round, confirmed facts, and the full relevant body; do not take over existing changes from other authors in the workspace. Silently exit for read-only, test/format, artifacts, third-party, or content without long-term semantic impact.
- Confirm real changes in responsibilities, contracts, objects, rules, state restoration, collaboration, resources, or source code anchors according to Scenario Closure; mechanical candidates are used only for localization.
- Perform minimal revisions on the original definitions specified in the Framework; do not append a changelog. Structural changes are only handled within the authorized scope of the current task, otherwise left for explicit
init/reinit/update. - Protect unrelated and human-authored content according to the Publication Contract, using read-time hashing, CAS, and read-back; when body content is written, run mechanical verification for affected domains.
explicit, symbol, path-only, and anchorReviewCandidates are not overrides or retirement conclusions. If there is no impact, do not output noop or run fixed commands.