Fix plain-files pull for Slack recent files

[mohammedahmed18]

Summary

• Include Slack recent.md pages in connector enumeration so plain-files pulls can materialize channel content.
• Let directory-target pulls work for plain-files projections, recursively discover connector directories, and hydrate all missing/stub child pages under an explicit directory pull.
• Skip refetching already-hydrated projected child pages during directory pulls, while still hydrating missing files and remaining stubs.
• Preserve plain-files remote moves during directory pulls by reusing the mount-root move plan/rename behavior, including recursive connector directories.
• Avoid rewriting existing stubs and avoid Notion-only missing-media repair for Slack pulls; update stub access polling so daemon rewrites do not look like user reads.
• Give daemon pull requests a pull-specific 2-hour default timeout so large Slack hydration pulls can finish without racing a direct fallback.

Test Plan

• cargo fmt --check
• cargo test -q -p loc-cli commands::tests::daemon_request_timeout_for_uses_pull_specific_default
• cargo test -q -p loc-cli --test pull
• cargo test -q -p localityd pull::tests
• cargo test -q -p locality-slack
• cargo build -q -p loc-cli -p localityd -p locality-slack
• cargo test --workspace --all-targets (rerun passed after one order-sensitive desktop test passed individually)
• git diff --check

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
locality	Ready	Preview, Comment	Jul 23, 2026 8:17pm

[github-actions]

Codex review

• major crates/localityd/src/pull.rs:992
  Plain directory pulls now hydrate every page child under non-database directories. That fixes Slack recent.md, but it also changes Google Docs folder pulls: after a root pull, loc pull <mount>/marketing will fetch and hydrate every stubbed Google Doc in that Drive folder, violating the existing online-only behavior for workspace folders and causing unexpected Docs API reads. Minimal fix: gate this auto-hydration to Slack or a connector capability, and add a Google Docs folder-pull test asserting nested docs remain stubs and docs.get_count() == 0.

• major crates/localityd/src/pull.rs:1611
  write_stub_if_needed now skips all existing page paths, including stale stub files. Previously, existing stubs were refreshed while non-stub local edits were preserved. With this change, a pull that sees updated remote title/version or connector-provided stub frontmatter updates the entity store but leaves the Markdown stub’s frontmatter stale; for row/property stubs this can expose old remote properties and later edits can be based on obsolete metadata. Minimal fix: only skip when the existing path is non-stub, or when the existing stub content exactly equals stub_markdown(entry). Add a test where an existing stub has old frontmatter and a pull rewrites it to the new stub without touching hydrated/non-stub files.