The orphan worktree gives historical notes a useful lifecycle without forcing a separate repository. One operational detail I would add is that .gitignore keeps .sdlc out of the product index, but does not by itself prevent every agent or filesystem search tool from reading that mounted directory. A recursive search that ignores Git's exclusions can still retrieve those obsolete requirements.
It would be useful to document the default search boundary alongside the mounting instructions, with a deliberate way to consult archived records when needed. Testing the actual agent's file-discovery behavior after setup would validate the context benefit more directly than counting tracked Markdown files alone.
The orphan worktree gives historical notes a useful lifecycle without forcing a separate repository. One operational detail I would add is that .gitignore keeps .sdlc out of the product index, but does not by itself prevent every agent or filesystem search tool from reading that mounted directory. A recursive search that ignores Git's exclusions can still retrieve those obsolete requirements.
It would be useful to document the default search boundary alongside the mounting instructions, with a deliberate way to consult archived records when needed. Testing the actual agent's file-discovery behavior after setup would validate the context benefit more directly than counting tracked Markdown files alone.