name: doc-sync description: Synchronizes documentation across agents to reflect planning changes and status updates.

Doc Sync Skill

Purpose: Automate document sync across agents so plan changes, progress, and open questions update in real time When to use: After Codex Validator completes, after Requirements Completion Check, before Documentation Finalize Outputs: {tasksRoot}/{feature-name}/context.md, pending-questions.md, flow-report.md

Goals

Problems

• Previous system: docs updated only by Documentation Agent
• No mid-step sync: Validator plan edits not reflected immediately
• Delayed feedback loop: Implementation Agent references stale context.md

Solution

• Update context.md immediately after Codex Validator
• Track real-time progress (flow-report.md)
• Centralize open questions (pending-questions.md)

Auto-trigger points

1. After Codex Validator completes

• Apply Validator auto_apply items to context.md
• Add new recommendations to pending-questions.md
• Mark "Planning complete" in flow-report.md

2. After Requirements Completion Check

• Add incomplete items to pending-questions.md
• Record "Implementation re-run" in flow-report.md

3. Before Documentation Finalize

• Final sync (ensure all docs are current)
• Clear pending-questions.md or mark as Resolved
• Mark flow-report.md as complete

Input format

YAML (recommended)

feature_name: batch-management
updates:
  - file: context.md
    section: Phase 1
    action: append
    content: "Strengthen date input validation: limit to past 30 days"
  - file: pending-questions.md
    action: add_question
    priority: MEDIUM
    content: "Should error messages use Toast?"
    context: "Improve user experience"
  - file: flow-report.md
    action: update_phase
    phase: Planning
    status: completed

Manual trigger (user direct)

doc-sync start: batch-management
  - context.md: update Phase 1
  - pending-questions.md: add 1 item
  - flow-report.md: Planning complete

Supported files and actions

1. context.md

Action: append (add to end of section)

file: context.md
section: Phase 1
action: append
content: "Strengthen date input validation: limit to past 30 days"

Result:

## Phase 1: Mock-based UI (1h)
1. Type definitions (15m)
2. Mock data (10m)
...

### Validator Feedback (added 2025-12-20)
- Strengthen date input validation: limit to past 30 days

Action: update (edit specific content)

file: context.md
section: "Risks and Alternatives"
action: update
old_content: "Probability: Medium"
new_content: "Probability: Low (API spec confirmed)"

Action: prepend (add to beginning of section)

file: context.md
section: "Target Files"
action: prepend
content: "WARN Validator recommendations applied (2025-12-20)"

2. pending-questions.md

Action: add_question (add a question)

file: pending-questions.md
action: add_question
priority: HIGH
content: "What is the allowed range for past dates?"
context: "Validator feedback: recommend limiting to past 30 days"
options:
  - 30 days
  - 60 days
  - 90 days
  - unlimited

Result:

## Pending Questions

### [HIGH] What is the allowed range for past dates?
- **Found at**: 2025-12-20 09:25
- **Context**: Validator feedback: recommend limiting to past 30 days
- **Options**:
  - 30 days
  - 60 days
  - 90 days
  - unlimited
- **Status**: pending

Action: resolve_question (resolve a question)

file: pending-questions.md
action: resolve_question
question_id: 1
resolution: "Decided on 30 days"
resolved_at: "2025-12-20 09:30"

Result:

### [HIGH] ~~What is the allowed range for past dates?~~ (resolved)
- **Decision**: 30 days
- **Resolved at**: 2025-12-20 09:30

Action: clear (remove all questions, for Finalize)

file: pending-questions.md
action: clear
archive: true

3. flow-report.md

Action: update_phase (update phase status)

file: flow-report.md
action: update_phase
phase: Planning
status: completed
timestamp: "2025-12-20 09:25"

Result:

| Phase | Status | Start | End |
|-------|--------|-------|-----|
| Planning | OK complete | 09:00 | 09:25 |
| Implementation | In progress | 09:30 | - |

Action: add_event (add event)

file: flow-report.md
action: add_event
event: "Validator feedback applied"
description: "Added date validation to context.md"
timestamp: "2025-12-20 09:25"

Result:

## Key Events

- [09:25] **Validator feedback applied**: added date validation to context.md

Output format

Success

OK Doc Sync complete

## Updated files
- context.md: added Validator feedback to Phase 1
- pending-questions.md: added 1 question (HIGH)
- flow-report.md: marked Planning complete

## Change summary
- Validator feedback: strengthen date validation (past 30 days)
- New question: decide past date range
- Planning phase complete (elapsed: 25m)

## Next steps
- Implementation Agent re-check (use latest context.md)
- Await pending-questions answers (HIGH 1 item)

Error

ERROR Doc Sync failed

## Error details
- context.md: section "Phase 1" not found
- pending-questions.md: update succeeded
- flow-report.md: file not found (needs creation)

## Action required
- Check context.md section structure
- Create flow-report.md manually

## Partial success
1/3 files updated

Related agents/skills

Inputs (consumed)

• Codex Validator: auto_apply, user_confirm items
• Moonshot Agent (Completion Check): incomplete_items list
• Documentation Agent: final sync request

Outputs (provided)

• Implementation Agent: latest context.md
• Moonshot Agent: pending-questions count, flow-report status
• Documentation Agent: confirmation that docs are up to date

File structure

Document paths

{tasksRoot}/{feature-name}/
|-- context.md              # implementation plan (real-time updates)
|-- pending-questions.md    # open questions (real-time updates)
`-- flow-report.md          # per-phase progress (real-time updates)

Archive (optional)

{tasksRoot}/{feature-name}/archives/
|-- context-v1.md              # before Validator feedback
|-- context-v2.md              # after Validator feedback
`-- pending-questions-resolved.md  # resolved questions

Usage scenarios

Scenario 1: Apply Codex Validator feedback automatically

Validator output:

status: pass_with_changes
auto_apply:
  - priority: HIGH
    target: context.md
    section: Phase 1
    content: "Strengthen date input validation: limit to past 30 days"
user_confirm:
  - priority: MEDIUM
    content: "Recommend changing error messages to Toast"

Doc Sync run:

feature_name: batch-management
updates:
  - file: context.md
    section: Phase 1
    action: append
    content: "### Validator Feedback (HIGH)
- Strengthen date input validation: limit to past 30 days"
  - file: pending-questions.md
    action: add_question
    priority: MEDIUM
    content: "Should we change error messages to Toast?"
    context: "Validator recommendation"
  - file: flow-report.md
    action: add_event
    event: "Validator feedback applied"
    description: "context.md updated + 1 pending question added"

Result:

• Implementation Agent reads latest context.md and applies date validation
• User answers pending-questions about Toast

Scenario 2: Re-run after Requirements Completion Check

Moonshot Agent Completion Check result:

status: incomplete
incomplete_items:
  - "Add error alert"
  - "Configure menu/permissions"

Doc Sync run:

feature_name: batch-management
updates:
  - file: pending-questions.md
    action: add_question
    priority: HIGH
    content: "Implement error alert handling"
    context: "Completion Check: missing in preliminary agreement"
  - file: pending-questions.md
    action: add_question
    priority: HIGH
    content: "Configure menu/permissions"
    context: "context.md Phase 3 checkpoint incomplete"
  - file: flow-report.md
    action: update_phase
    phase: Implementation
    status: "re-run required"

Result:

• Re-run Implementation Agent (only incomplete items)
• flow-report.md records the re-run

Scenario 3: Final sync before Documentation Finalize

Documentation Agent request:

feature_name: batch-management
updates:
  - file: context.md
    section: "Final State"
    action: append
    content: "- [x] All phases complete
- [x] Verification passed
- [x] Documentation complete"
  - file: pending-questions.md
    action: clear
    archive: true
  - file: flow-report.md
    action: update_phase
    phase: Documentation
    status: completed

Result:

• pending-questions.md cleared (resolved items moved to archives)
• flow-report.md marked complete
• context.md updated with final state

Tips

1. Avoid conflicts

• Detect simultaneous updates by comparing timestamps
• Notify user and resolve manually on conflict

2. Versioning (optional)

• Backup prior versions for important changes
• Store as context-v1.md, context-v2.md

3. Rollback support

• Save snapshot before last change
• Restore immediately on failure

4. Verification automation

• Validate file structure after updates
• Check required sections are present

Expected impact

Qualitative impact

1. Real-time feedback loop: Validator -> Doc Sync -> Implementation (immediate reflection)
2. Document consistency: all agents reference the latest docs
3. Centralized open questions: pending-questions.md as the source of truth
4. Progress transparency: real-time tracking via flow-report.md

Quantitative impact

• Feedback turnaround: manual 10m -> instant (100% reduction)
• Doc mismatch errors: 30% -> 0% (eliminated)
• Rework prevention: instant feedback saves ~15m on average

Implementation details

Quality bar

• Preserve existing content (do not change section structure)
• Prevent conflicts (detect concurrent updates)
• Ensure atomicity (rollback on partial failure)
• Automate validation (check structure after updates)

Error handling

1. Missing file: auto-create (use template)
2. Missing section: warn and append to end
3. Concurrent update conflict: compare timestamps and notify user
4. Format error: validation fails -> rollback

Logging

• Record all updates in flow-report.md automatically
• Include timestamp, changed files, change details
• Store detailed logs on failure

Enabling this skill keeps all docs in real-time sync.