Skills

All 9 public skills in the vantage-peers plugin v2.4.0 — descriptions, trigger phrases, when to use, and minimal examples.

Skills

Skills are reusable workflow protocols that encode VP best practices. Each skill is invoked by a trigger phrase (natural language) or its corresponding slash command. Skills use VP MCP tools internally and enforce patterns like Evidence-Bound Done and T-VERIFY doctrine.

All 9 Skills

check-messages

Description: Poll unread messages from other orchestrators, respond to any that require action, and (in autonomous mode) auto-pick the next unblocked todo task.

Trigger phrases: "check messages", "any messages", "inbox", "peers", "new messages"

When to use:

• At the start of every session to check for dispatched work
• When running autonomously to chain tasks (the skill self-chains via Step 6)
• When a peer orchestrator may have sent instructions or completed delegated work

Example:

User: check messages

The skill detects your orchestrator mode (human vs autonomous), polls check_messages, displays any unread messages with their senders, responds to any that require action, marks them as read, and in autonomous mode picks the next priority task.

check-tasks

Description: Fetch all tasks assigned to your orchestrator role, filter out done tasks, sort by priority, and flag any blocked tasks.

Trigger phrases: "my tasks", "task list", "what should I work on", "backlog"

When to use:

• To get an overview of your current workload
• Before starting a session to know what's queued
• When you want to see blocked tasks and their blockers

Example:

User: what should I work on today?

The skill calls list_tasks with your assigned role, groups by priority (urgent → high → medium → low), surfaces blocked tasks with their dependency IDs, and presents the next unblocked task.

close-day

Description: End-of-day wrap routine — updates open task statuses, writes a diary entry, stores a session summary as memory, and calls set_summary to closed.

Trigger phrases: "close day", "end of day", "wrap up", "close session"

When to use:

• At the end of a working session before stopping Claude Code
• Before a planned context compaction

Example:

User: close day

The skill prompts for any outstanding updates, writes a diary entry for today, stores a reference memory with session highlights, and sets your orchestrator summary to closed/standby.

daily-start

Description: Morning session start — loads VP context (recent memories, active tasks, messages), presents the day plan for human operators or auto-picks the highest-priority task for autonomous orchestrators.

Trigger phrases: "start the day", "morning plan", "daily planning", "session start"

When to use:

• At the beginning of every working session
• When resuming after a context compaction

Example:

User: start the day

In human mode: recalls recent project memories, lists active tasks, checks messages, and presents a proposed session plan. In autonomous mode: directly picks and starts the top-priority unblocked task.

pre-compact

Description: Session snapshot before context compaction — saves full session state (active missions, tasks, blockers, 3-line summary) as a reference memory and a briefing note.

Trigger phrases: "save context", "before compaction", "snapshot session"

When to use:

• When Claude Code warns that context is approaching the limit
• Before intentionally compacting to continue work in a fresh context

Example:

User: save context before compaction

The skill calls store_memory with a snapshot of current session state and create_briefing_note with a structured handoff note, so the next session can recall exactly where things were left.

recall

Description: Semantic + BM25 hybrid search across VP memories. Auto-detects the most likely namespace from the query.

Trigger phrases: "recall", "search memory", "what do we know about", "look up"

When to use:

• Before answering any factual question about project state, history, or decisions
• When you need to find a previously stored fix pattern, spec, or decision

Example:

User: recall what we decided about the auth architecture

The skill constructs a hybrid recall query (semantic + BM25), searches across the relevant namespace (auto-detected from query context), and returns the top matching memories with their types and namespaces.

standup

Description: Generate a structured standup report (DONE / IN PROGRESS / BLOCKERS / GIT sections) and file it as a briefing note.

Trigger phrases: "standup", "status report", "daily report", "sitrep"

When to use:

• Daily standup or shift handoff
• When a team coordinator needs a structured status update
• Before a planning meeting

Example:

User: standup

The skill calls list_tasks for recent completions and in-progress items, checks git for recent commits (if Bash is available), assembles the 4-section report, and calls create_briefing_note with topic standup.

vantage-peers-init

Description: Verify VP setup — checks MCP registration, tests /health, and smoke-tests auth via recall. Produces a PASS/FAIL report with specific fix instructions per failure.

Trigger phrases: "verify VP setup", "VP smoke test", "init vantage-peers"

When to use:

• After initial plugin installation
• After changing .mcp.json or bearer secret
• After a Railway redeployment

Example:

/vantage-peers-init

All 3 checks must PASS before using other skills. If any fail, follow the specific fix instruction output by the skill.

write-diary

Description: Write a structured diary entry — asks one grounding question, then constructs an entry with highlights, blockers, and a reflection section.

Trigger phrases: "write diary", "diary entry", "log today", "journal entry"

When to use:

• At the end of a meaningful session
• After completing a significant milestone
• As part of the close-day skill (which calls this internally)

Example:

User: write diary

The skill asks "What was the most important thing that happened today?" then calls write_diary with a structured entry covering highlights, what was learned, and any open blockers.