vantageos-agency
OrganizationOpen-source coordination backend for AI agent teams. Shared memory, messaging, tasks via MCP. 20 tables, 75 tools. Convex-native.
Categories
Indexed Skills (28)
dispatch-subagent
Compose and dispatch a Claude Code subagent (the `Agent` tool) with the correct brief template marker pre-injected so the `enforce-brief-template` hook never blocks. Use this skill whenever the user says "dispatch a subagent", "agent for X", "spawn agent", "delegate to subagent" — even if they don't say "dispatch-subagent" explicitly.
check-messages
Check and respond to peer messages from other orchestrators, and (in autonomous mode) also list + pick the next unblocked todo task. Use this skill whenever the user says "check messages", "read messages", "any messages", "peers", "inbox", "new messages" -- even if they don't say "check-messages" explicitly.
check-tasks
List the tasks assigned to the current orchestrator, sorted by priority and dependency-aware, while keeping the response under the 60 KB cap by always requesting the lite projection with a default limit. Use this skill whenever the user says "check tasks", "my tasks", "what tasks", "pending tasks", "task list", "todo list", "what should I work on", "backlog" — even if they don't say "check-tasks" explicitly.
triage-issue
Auto-create a mission with IRP T0-T7 tasks from a GitHub issue. Use this skill whenever the user says "triage issue", "create mission for issue", "handle this issue", "treat this issue", shares a GitHub issue URL, or when a new issue needs to be assigned to an orchestrator -- even if they don't say "triage-issue" explicitly.
write-diary
Write a daily AI Diary entry in Pi's voice for perfectaiagent.xyz. Use this skill whenever the user says "diary", "write diary", "daily entry", "Pi's diary", "journal entry", "day entry", or asks to document what happened today — even if they don't say "diary" explicitly.
briefing-write
Wrap create_briefing_note and update_briefing_note with topic taxonomy enforcement, participants list normalization, content size pre-flight, and auto-linking of related taskIds / missionIds mentioned in the content. Use this skill whenever the user says "write briefing note", "briefing", "decision note", "create briefing", "save briefing" -- even if they don't say "briefing-write" explicitly.
dispatch-message
Pre-format every outbound peer message into the v2 telegraphic grid so it satisfies the fleet hooks (no-task-in-message marker, signature, ship-24-7, no time-estimates) on the first try AND removes the narrative slot that invites chronological prose. Use this skill whenever the user says "send a message", "DM <orch>", "broadcast", "reply to <orch>", "tell <orch> X" -- even if they don't say "dispatch-message" explicitly.
dispatch-task-complete
Close a VantagePeers task with an auto-assembled proof-token completionNote so the evidence-bound-done hook never blocks. Use this skill whenever the user says "complete task <id>", "mark done", "close <id>", "finish task", "task done", "wrap task", "close task", or asks to move a task to done/review -- even if they don't say "dispatch-task-complete" explicitly.
dispatch-task-create
Create a VantagePeers task for another orchestrator with a properly-formed description (VERIFICATION + TESTS + IRP blocks + delegation triplet on batch ops + auto allow-no-notify marker on multi-task dispatch) so the enforce-task-quality + enforce-pi-task-doctrine + enforce-create-task-notify-followup hooks never block. Use this skill whenever the user says "create task for <orch>", "dispatch task", "queue work for <orch>", "ask <orch> to do X" — even if they don't say "dispatch-task-create" explicitly.
dispatch-task-start
Wrap `mcp__vantage-peers__start_task` with an automatic stale-in_progress sweep so the `enforce-irp-sequence` hook never blocks the caller. Use this skill whenever the user says "start task <id>", "begin task", "pick up <id>", "start_task" — even if they don't say "dispatch-task-start" explicitly.
fix-pattern-cycle
Single skill covering the full VantagePeers fix-pattern lifecycle — create a pattern from a recurring incident, log fix attempts, validate fixes with proof, link patterns to issues, and search/list existing patterns with envelope-safe defaults. Use this skill whenever the user says "fix pattern", "create fix pattern", "add fix attempt", "validate fix", "link issue to pattern", "search fix patterns", "list fix patterns" — even if they don't say "fix-pattern-cycle" explicitly.
messages-history
Browse VantagePeers peer-message history with day/sender filters, envelope-safe defaults, broadcast read-receipt audits, and safe mark-as-read / delete operations, so orchestrators can audit conversations without the unread-inbox tunnel vision of check-messages. Use this skill whenever the user says "messages history", "show past messages", "messages from <peer>", "messages on day N", "broadcast status", "who read the broadcast", "mark messages read", "delete a message" -- even if they don't say "messages-history" explicitly.
mission-bootstrap
Bootstrap a VantagePeers mission with the full IRP task chain (plan → execute → verify → ship) wired with dependencies in a single skill invocation. Use this skill whenever the user says "bootstrap mission", "start a mission for X", "scaffold IRP for X", "create mission with tasks" — even if they don't say "mission-bootstrap" explicitly.
open-pr
The only sanctioned way to open a pull request. In one invocation it does three atomic acts: (1) runs gh pr create with the fixed marker # via-open-pr appended, (2) creates exactly one reviewer (Eta) review task citing PR number, head SHA, and repo, (3) notifies channel eta,pi with the v2 telegraphic grid. Use this skill whenever the user says "open a PR", "create a PR", "raise a pull request", "gh pr create" — even without naming open-pr explicitly.
review-pr
The reviewer's routine for rendering a verdict on a GitHub pull request. Use this whenever you are asked to "review PR", "gate PR", "render a verdict", "re-gate", or issue an APPROVED / REVISE on a pull request — even if the words "review-pr" are not said. It mechanizes the flow rule "update before verdict": the reviewer brings the PR up to the current base BEFORE gating, then gates the head it OBTAINED, so a verdict is never invalidated by a base that moved.
close-day
End-of-day routine: close tasks with their time, leave the repository and the disk clean, write the diary, harvest friction, and hand over to tomorrow. Use this skill whenever the user says "close day", "end of day", "fin de journée", "bonne nuit", "wrap up", "call it a day", "close session", "daily close", or mentions ending their work session -- even if they don't say "close-day" explicitly.
daily-start
This skill should be used when the user asks to "start the day", "morning plan", "daily planning", "what's on my plate today", "plan today", "session start", "daily start", or mentions wanting to organize their day or review pending tasks -- even if they don't say "daily-start" explicitly.
pricing-research
Process reproductible de recherche pricing pour tout produit/service ElPi Corp. 4 axes parallèles avec gate approval competitors. Subagent B = 2 phases (candidates list → approval Laurent → deep scrape Firecrawl). Output multi-format pour comparison table landing + articles blog "Notre app vs competitor".
request-gate
Force any gate/review request for an artifact NOT versioned in git (a gitignored hook, an adversarial probe/sonde, a local one-off script, or an announced-but-not-transported VR content pair) to transport the replayable artifact itself via create_briefing_note — never just ratios + a sha256. A reviewer cannot refute what they cannot replay. Use this skill whenever the user says "gate this", "send for review", "ask Eta/Omega/Pi to review", "request a verdict on X" -- even if they don't say "request-gate" explicitly, whenever the artifact in question lives outside git.
standup
Daily standup report -- what was done, what's in progress, blockers, git status. Use this skill whenever the user says "standup", "status report", "daily report", "what did you do", "progress report", "sitrep", "report", "where are you", or asks about current work status -- even if they don't say "standup" explicitly.
recall
Search VantagePeers for stored memories. Use this skill whenever the user says "recall", "remember", "what do we know about", "search memory", "look up", "find in memory", or asks about past decisions, context, or stored knowledge -- even if they don't say "recall" explicitly.
convex-create-component
Designs and builds Convex components with isolated tables, clear boundaries, and app-facing wrappers. Use this skill when creating a new Convex component, extracting reusable backend logic into a component, building a third-party integration that owns its own tables, packaging Convex functionality for reuse, or when the user mentions defineComponent, app.use, ComponentApi, ctx.runQuery/runMutation across component boundaries, or wants to separate concerns into isolated Convex modules.
convex-migration-helper
Plans and executes safe Convex schema and data migrations using the widen-migrate-narrow workflow and the @convex-dev/migrations component. Use this skill when a deployment fails schema validation, existing documents need backfilling, fields need adding or removing or changing type, tables need splitting or merging, or a zero-downtime migration strategy is needed. Also use when the user mentions breaking schema changes, multi-deploy rollouts, or data transformations on existing Convex tables.
convex-performance-audit
Audits and optimizes Convex application performance across hot-path reads, write contention, subscription cost, and function limits. Use this skill when a Convex feature is slow or expensive, npx convex insights shows high bytes or documents read, OCC conflict errors or mutation retries appear, subscriptions or UI updates are costly, functions hit execution or transaction limits, or the user mentions performance, latency, read amplification, or invalidation problems in a Convex app.
convex-quickstart
Initializes a new Convex project from scratch or adds Convex to an existing app. Use this skill when starting a new project with Convex, scaffolding with npm create convex@latest, adding Convex to an existing React, Next.js, Vue, Svelte, or other frontend, wiring up ConvexProvider, configuring environment variables for the deployment URL, or running npx convex dev for the first time, even if the user just says "set up Convex" or "add a backend."
convex-setup-auth
Sets up Convex authentication with user management, identity mapping, and access control. Use this skill when adding login or signup to a Convex app, configuring Convex Auth, Clerk, WorkOS AuthKit, Auth0, or custom JWT providers, wiring auth.config.ts, protecting queries and mutations with ctx.auth.getUserIdentity(), creating a users table with identity mapping, or setting up role-based access control, even if the user just says "add auth" or "make it require login."
friction-digest
Weekly aggregator of fleet friction memories. Reads VantagePeers `audit/friction` namespace last 7 days, ranks patterns by frequency, identifies top 10, auto-creates improvement missions for top 3 with clear fix path, generates briefing note for Pi+Laurent. Triggers: weekly cron Sunday 22:30 `/friction-digest`, manual 'friction digest', 'weekly friction harvest', 'friction review'.
pi-merge-fleet-pr
Pi-signed merge of an Eta-APPROVED fleet client-facing PR — wraps the 7-hook gauntlet (enforce-pi-authorization + enforce-vr-consult + enforce-create-task-notify-followup + enforce-friction-field) into one atomic skill that passes first-try.
Bio shown is the top-scored skill's repo description as a fallback — real GitHub bios land in a future update.