CalDAVCalDAV同步

When to Use

Use when the user needs to work with CalDAV calendars through a local vdirsyncer + khal stack, especially for iCloud, Fastmail, Nextcloud, DAViCal, Radicale, or other standards-based calendar servers.

This skill is for local-first querying, event creation, safe edits, troubleshooting stale sync state, and handling ambiguous matches without corrupting recurring events or overwriting the wrong calendar.

Requirements

• - vdirsyncer and khal must be installed and available in PATH.
• The CalDAV account and collection config must already exist outside this skill.
• A TTY is required for interactive khal edit workflows.

Core Rules

1. Sync is part of every real operation

• - Treat vdirsyncer sync as part of the workflow, not an optional cleanup step.
• Sync before reads when freshness matters, and sync again after confirmed writes so local and remote state converge.
• If collections are missing or a server path changed, use vdirsyncer discover before assuming the calendar is empty.
• Do not trust a stale khal result until sync and cache state have been checked.

2. Use bounded windows and explicit calendar scope

• - Prefer day, 7-day, 14-day, or exact date range queries over open-ended searches.
• Narrow to a specific calendar whenever the user already knows the target calendar.
• Resolve ambiguous phrases such as "next Friday" or "this evening" into exact dates, times, and timezone assumptions before writing.
• Titles are not unique identifiers, so duplicate-event risk rises quickly when searching across every calendar at once.

3. Respect khal's editing limits

• - khal edit is interactive and needs a TTY, so do not present it as a non-interactive batch primitive.
• INLINECODE10 has only rudimentary recurrence editing and cannot edit event timezones directly, so complex recurring or timezone-sensitive events need extra caution.
• For fragile recurring series, DST-sensitive events, or uncertain matches, prefer inspect-first and recreate-only-with-approval over aggressive in-place edits.
• If the user wants bulk recurring surgery, stop and explain the risk before touching anything.

4. Verify every write with a read-back pass

• - Before changing an event, read the target window first so the agent sees the exact current state.
• After create, update, or delete, run a read-back check in the same bounded window and report the returned title, time, and calendar.
• Use title plus date/time plus calendar, or UID when available, to confirm the right event changed.
• If verification is ambiguous or inconsistent, stop and surface the conflict instead of claiming success.

5. Treat local vdir state and conflict policy as real data

• - vdirsyncer is synchronizing real local .ics state, not just acting as a remote viewer.
• The configured conflict_resolution policy can overwrite one side, so do not assume an "a wins" or "b wins" setup is harmless.
• Manual filesystem edits, cache resets, or storage path changes should be deliberate and reversible.
• Deleting khal's cache database is a troubleshooting move for stale cache behavior, not a default fix for every mismatch.

6. Protect connection details, URLs, and certificates

• - Confirm the CalDAV base URL and collection path before debugging deeper issues.
• Certificate and TLS errors are blockers; stop and fix the trust chain before continuing.
• Keep private connection details and sync config out of summaries unless the user explicitly asks for them.

7. Finish with operational clarity

• - Every answer should end with the exact calendar scope, time window, action taken or proposed, and whether another sync is needed.
• If blocked, name the real blocker precisely: missing vdirsyncer, missing khal, missing TTY, undiscovered collections, login failure, or ambiguous event match.
• If the safest action is read-only, say so directly instead of improvising a write path.

Common Traps

• - Querying without syncing first -> stale answers and wrong scheduling decisions.
• Editing by title only -> the wrong duplicate event gets changed or deleted.
• Treating recurring events like normal one-offs -> series corruption or DST drift.
• Searching every calendar by default -> noisy matches and accidental writes to the wrong calendar.
• Using a one-sided conflict_resolution policy blindly -> local or remote data loss.
• Deleting khal.db too early -> symptoms disappear briefly while the real sync bug remains.
• Assuming any WebDAV-looking URL is a valid CalDAV calendar collection -> discovery and auth failures.
• Reporting success before read-back verification -> hidden mismatch between local cache and remote server.

Related Skills

• - calendar-planner - Plan weeks, repair conflicts, and turn calendar state into defended scheduling decisions.
• INLINECODE21 - Handle general scheduling requests when the user needs timing help beyond raw CalDAV operations.
• INLINECODE22 - Use provider-specific Fastmail mail and calendar APIs when CalDAV is not enough.
• INLINECODE23 - Turn calendar deadlines and follow-ups into reminder workflows.

Feedback

• - If useful: INLINECODE24
• Stay updated: INLINECODE25