em%3aupdate（em%3a 更新）

Compatibility: If AskUserQuestion is unavailable, present options as a numbered list and wait for the user's reply. If Task is unavailable, run parallel steps sequentially. The context: fork and agent: frontmatter fields are Claude Code-specific — on OpenCode and VS Code Copilot they are ignored and the skill runs inline using the current model.

Update an EM plan or task milestone file. Two input modes:

• Annotations: > and >> markers already written in plan files
• Conversational: plain-language description of what to change

Annotation syntax

> instruction — standalone line, applies to the block immediately below it:

> rewrite this phase for the new auth approach
### Phase 2 - Authentication Migration

content >> instruction — inline suffix, applies to that line:

- Migrate user sessions >> mark done
- Add refresh token rotation >> remove this task
- Configure OAuth providers >> change to use env vars

Both can appear in plan.md and any tasks/*.md.

Step -1: Parse Flags

BG_MODE = true if --bg present
SILENT  = true if --silent present

If BG_MODE=true: skip the confirmation in Step 1 and send a desktop notification after Step 5.

Step 0: Select Plan

Check for plan slug argument. If not provided:

1. List .codevoyant/em/plans/*/plan.md sorted by modification time (most recent first)
2. If only one plan, auto-select it
3. If multiple, use AskUserQuestion to present the list and ask the user to choose
4. If none exist, inform user to run /em:plan first

Verify .codevoyant/em/plans/{slug}/plan.md exists. Set PLAN_DIR=".codevoyant/em/plans/{slug}".

Step 0.5: Determine Input Mode

Check the argument string and triggering message for a change description:

• If a non-slug argument is present (e.g., /em:update add observability to phase 3), treat everything after the slug as CHANGE_DESCRIPTION
• If neither: CHANGE_DESCRIPTION is empty -> annotation mode

Set INPUT_MODE:

• conversational — CHANGE_DESCRIPTION is non-empty
• annotations — scan plan files for > / >> markers
• If both present, process conversational change first, then apply any annotations

Step 1: Process Conversational Change (if INPUT_MODE includes conversational)

Read plan.md and any relevant tasks/*.md to understand current structure.

Translate CHANGE_DESCRIPTION into concrete edits:

1. Identify exactly which files are affected
2. Determine what needs to change -- specific new text, removed lines, renamed sections
3. Show a concise preview before applying:

Proposed changes for: "{CHANGE_DESCRIPTION}"

  plan.md
    + Phase 2: add observability milestone under deliverables

  tasks/develop.md
    + Task: "Set up distributed tracing with OpenTelemetry"
    + Failure mode: "Trace context lost across service boundaries"

Apply these changes?

If the change marks a task as done, after applying offer: "Push status update to Linear? (uses mcp__linear-server__save_issue with completed state)"

If BG_MODE=true, auto-apply without asking.

Otherwise use AskUserQuestion:

question: "Apply these changes to {slug}?"
header: "Roadmap Update"
multiSelect: false
options:
  - label: "Apply"
    description: "{first line of change summary}"
  - label: "Adjust"
    description: "Let me clarify what I want"
  - label: "Cancel"
    description: "Don't change the plan"

• Apply: proceed
• Adjust: ask "What should be different?", update proposed changes, re-confirm
• Cancel: exit

Step 2: Scan for Annotations

grep -rn "^>" {PLAN_DIR}/plan.md {PLAN_DIR}/tasks/ 2>/dev/null
grep -rn ">>" {PLAN_DIR}/plan.md {PLAN_DIR}/tasks/ 2>/dev/null

For each annotation, parse: FILE, LINE_NUM, CONTENT (before >>), INSTRUCTION.

If INPUT_MODE=annotations and no annotations found:

No annotations found in {slug}.

To annotate, edit plan.md or tasks/*.md directly:
  > rewrite this section for the new approach     -- applies to next block
  - task name >> mark done                        -- applies to this line

Exit.

Step 3: Apply Each Annotation

Work bottom-to-top within each file so line numbers stay valid.

Remove the annotation marker after applying. Log each change for the summary.

Step 4: Consistency Pass

After all changes:

• Verify task milestone files still exist: {PLAN_DIR}/tasks/design.md, tasks/develop.md, tasks/deploy.md
• Check milestone naming and task numbering is consistent throughout plan.md
• Verify "NOT this period" section still accurately reflects deferrals

Step 5: Validation Pass

Run 2 validation rounds autonomously — no user prompts.

For each round, launch parallel agents (model: claude-haiku-4-5-20251001, run_in_background: true):

Plan-level agent -- checks plan.md: milestones have objectives/deliverables/risks, assumptions listed.

Per-task-file agents -- one per tasks/*.md file touched in this update: each task has requirements, ACs, design/SA fields filled; tasks are specific and actionable.

Collect results (TaskOutput block: true). Auto-fix any NEEDS_IMPROVEMENT issues. Run round 2 after fixes. Cap at 3 rounds.

Step 6: Report

Updated: {slug}

  Changes applied:
    plan.md:24           -- added observability milestone
    tasks/develop.md:8   -- added OpenTelemetry task

  Validation: {N} rounds -- {PASS | X issues remain}

If an annotation was ambiguous or could not be applied:

⚠️  Skipped annotation at {file}:{line}: {reason}
    Annotation preserved — resolve manually.

Step 7: Notify (--bg only)

If BG_MODE=true and SILENT=false:

npx @codevoyant/agent-kit notify --title "em:update complete" --message "Roadmap '{slug}' updated"