Safe Section Deletion

Deleting content that looks redundant or leftover can silently break something that referenced it: an anchor link, an import, a route. This is the discipline that catches that before it happens, not after.

Procedure

1. Before deleting, identify what could reference the thing being removed:

• An HTML section or id → anchor links (#section-id) from anywhere in the site, including other pages.
• A named export/component → imports across the codebase.
• A route or page → nav links, redirects, sitemap entries.
• A CSS class → other stylesheets or inline styles depending on it.

2. Grep for every plausible reference form before touching anything:

grep -rn "#section-id" .        # anchor links
grep -rn "ComponentName" .      # imports/usages
grep -rn "/route-path" .        # route references

Check for the id/name used both as an href target and as a raw string (some references are constructed dynamically rather than hardcoded).

3. Zero in-repo references confirmed → safe to delete. Any hit → resolve it first (update the reference or confirm it's dead too) before deleting the target.

4. For actual code symbols (functions, classes, methods) rather than markup, prefer a proper call-graph-aware impact tool if one is available in the project (e.g. an impact-analysis MCP tool) over a plain grep. A grep catches text matches, not semantic call relationships, and can both over- and under-count real usages (a same-named local variable creates a false positive; a re-exported symbol under a different local name creates a false negative).

5. If the content might be linked from outside the repo (an external site linking to an anchor, a bookmarked URL, an API consumer), say so explicitly. A repo-wide grep can't see that, and the user may need to consider it separately (e.g. a redirect) before the deletion is truly safe.

Done-when

Every plausible reference form has been checked and is either confirmed absent or resolved, and desktop+mobile (for web content) is verified after the deletion, not just before.