personal-chief-of-staff

name	personal-chief-of-staff
description	Use when the user wants to complete a daily journal or wind down, requests a daily chief-of-staff review, requests a weekly or quarterly review, later revisits, resumes, approves, edits, defers, skips, or otherwise decides visible chief-of-staff actions, or another workflow requests current cross-source chief-of-staff context. Do not use for isolated task creation, issue writing, email processing, calendar editing, health analysis, meeting preparation, or project planning.
license	MIT
compatibility	Requires access to the user's chosen authoritative sources. Obsidian workflows require a running Obsidian app and its CLI.

Personal Chief of Staff

Turn live personal and work evidence into an interactive review that helps the user orient, act, reflect, and learn. The user's existing systems remain authoritative; this skill adds no database, run ledger, cache, or generated brief archive.

Resolve visible action responses first

Before selecting a mode, determine whether the current message approves, edits, defers, skips, declines, revisits, or otherwise decides actions from a visible chief-of-staff bundle. Resolve the response against that exact bundle and its originating mode. Do not infer an action from a bare number when the visible context does not identify it.

For each decided action, use its displayed identity, target, effect, and approval state. Read both shared resources, references/source-behavior.md and assets/review-bundle.md, plus the originating mode reference, then follow their existing shared and mode-specific application rules. For an approved Person-note, relationship Task, CRM-derived unrelated work, or CRM-derived writing-backlog effect, use the available managing-personal-crm companion in embedded mode and follow its references/applying-approved-actions.md semantics. Route CRM-derived unrelated work through the caller's configured canonical task or issue workflow and exact displayed destination; route a writing-backlog effect only through the configured canonical writing workflow and exact displayed destination. For either, search that destination for complete-meaning equivalence immediately before mutation. Report an equivalent that has appeared as Already satisfied; otherwise apply once and read the exact target back through the same authoritative interface. The chief-of-staff workflow keeps the action number, approval flow, result, and completion state; the companion creates no nested bundle. If the companion or the required canonical workflow, exact destination, search, write, or readback path is unavailable or ambiguous, report that relationship-derived action Manual rather than applying it under generic source rules or another destination.

For CRM-derived communication text, keep the exact displayed text in the conversation only. Never send it, create a draft, or create another artifact. If approved unchanged, report the chief-of-staff action Already satisfied because the editable text is already visible. An edit produces a revised proposal under the same chief-of-staff action number and requires a new exact approval. Keep its result and completion state with this workflow; do not use a generic mutation fallback or let the companion create a nested bundle.

An action-only response does not run new review discovery or prepare another review. It still performs every immediate pre-write target and destination re-read, equivalence, drift, dependency, and post-write readback check required by the loaded application rules. Its answer-first action-result narrative stays separate from the Source Access Audit, which reports only current target or destination reread and verification readback access. It does not repeat access from the originating bundle.

If the same message both decides visible actions and requests a new review, finish the visible action decisions first. Then select the requested mode and run its discovery as a separate read-only phase. Do not use newly retrieved evidence to reinterpret the earlier decisions. Render one Source Access Audit. The paragraph distinguishes Action access from Review discovery.

If the same message both decides visible actions and requests current cross-source context without a review mode, also finish the visible action decisions first. Then use the non-mode path below as a separate read-only context-discovery phase; do not open Wind-down, Weekly, or Quarterly, and do not let the new context reinterpret the earlier decisions. Render one Source Access Audit. The paragraph distinguishes Action access from Context discovery.

Completion: every visible action decision was resolved against its exact originating bundle and mode before any separately requested review or non-mode context discovery began.

Supply cross-source context without a review mode

When another workflow explicitly requests current cross-source chief-of-staff priority or context—and the message does not request wind-down, weekly, or quarterly review—do not select a review mode. If that message also decides visible actions, enter this path only after resolving them under the phase boundary above. Read references/source-behavior.md and assets/review-bundle.md. Retrieve only the evidence needed for the caller's judgment, distinguish fact from inference, and return priority context in the conversation with the answer-first Source Access Audit. The calling workflow retains ownership of its narrower operation. Do not open Wind-down, Weekly, or Quarterly, and do not invent a Morning path.

Before returning, always render the current response's Source Access Audit after the priority context, even when every requested read succeeds and no action is proposed. A cross-source response is incomplete unless that audit names every relevant role in the short paragraph specified in assets/review-bundle.md.

Completion: the caller has usable cross-source judgment without a chief-of-staff mode run or unrequested durable writes, and can see which current source reads did and did not succeed.

Select the mode

Choose exactly one mode from the request:

• Wind-down: The user asks to close the day, complete the daily journal, reflect, prepare tomorrow, run a daily chief-of-staff review, or invokes the scheduled wind-down mode. Generic daily review wording without weekly or quarterly context selects Wind-down.
• Weekly: The user asks to complete or discuss a weekly review.
• Quarterly: The user asks to complete or discuss a quarterly review.

An explicit mode wins. When the request does not identify one of these review contexts and is not a cross-source context request above, leave it with the narrower workflow that owns it.

Read both shared resources for every mode:

• references/source-behavior.md
• assets/review-bundle.md

Then read only the selected mode reference:

• Wind-down: references/wind-down.md
• Weekly: references/weekly.md
• Quarterly: references/quarterly.md

Completion: one mode and all required shared and mode-specific resources are selected before source retrieval begins, or the cross-source non-mode path was used instead.

Collaborate on judgment

Continue interactively rather than publishing a report and ending. The agent may retrieve, organize, compare, and draft substantial objective content. The user owns or explicitly approves subjective meaning, causal lessons, strategic judgment, and central published thinking.

Use the shared source behavior and review-bundle shape throughout the conversation. Present proposed external changes as one bundle with independently approvable actions.

Completion: the user has had a clear opportunity to correct the review and to approve, edit, defer, or skip every proposed action independently.

End explicitly

End the run as exactly one of:

• Complete: The review and all approved actions finished.
• Nothing material: The evidence supported the review and no attention or action was warranted.
• Partial: A useful review completed with named evidence limits.
• Unable to prepare reliably: Evidence was insufficient for a trustworthy review.
• Paused: The user intends to continue later.
• Skipped: The user chose not to conduct the review.

When resuming, refresh time-sensitive evidence. Resume in the same conversation when available; otherwise reconstruct from canonical sources, disclose that uncommitted conversational input is unavailable, and ask only for the missing human judgment.

Close with a short recap of what changed and what remains unapplied.

Completion: the ending is explicit, accurate, and consistent with the durable artifacts that now exist.