This repository also configures its own agents. See what cognia-next tells them →
Skill Claude CodeCodex
Use when a request needs triage or containment for stuck Cognia agents, failed tasks, runaway teams, runtime incidents, or goal failures; do not use when handling routine lifecycle, coordination, or healthy-run control, where cognia-host-agents applies.
Skill Claude CodeCodex
Use when handling routine Cognia agent lifecycle, coordination, approvals, interruptions, teams, fleets, or goals; do not use when diagnosing or containing an unhealthy run, where cognia-host-agent-incident applies.
Skill Claude CodeCodex
Use when a request concerns Cognia workflow runs, schedules, backfills, background jobs, monitors, or automation consent; do not use when the request is a one-off task workspace or agent-runtime operation.
Skill Claude CodeCodex
Use when a request needs deliberate Cognia backup export, verification, import, restore, disaster recovery, or post-recovery checks; do not use when handling routine system status, logs, sync, or secret administration.
Skill Claude CodeCodex
Use when a request combines Cognia connector health or target resolution with verified notification, integration, or Lark delivery; do not use when handling routine configuration, ingress registration, or status inspection without delivery.
Skill Claude CodeCodex
Use when inspecting routine Cognia connector health, adapters, attachments, WebSockets, integration ingress, or drafts; do not use when the task culminates in outbound delivery, where cognia-host-connector-delivery applies.
Skill Claude CodeCodex
Use when a request concerns Cognia workspace files, terminals, browser automation, code-server proxies, language servers, or read-only Git discovery; do not use when Git mutation, conflict handling, or recovery is required, where cognia-host-safe-git applies.
Skill Claude CodeCodex
Use when a Cognia plugin, skill, MCP, or provider rollout spans validation, installation, activation, diagnostics, rollback, or removal; do not use when handling a single lookup, permission change, runtime call, or routine diagnostic.
Skill Claude CodeCodex
Use when a request is a single Cognia plugin, skill, MCP, or provider lookup, diagnostic, permission, runtime, or lifecycle operation; do not use when staging and validating a multi-step release, where cognia-host-extension-rollout applies.
Skill Claude CodeCodex
Use when a request concerns Cognia memory retrieval, digital-twin sources or profiles, ingestion jobs, OCR extraction, or model management; do not use when the primary task is connector export or host-storage recovery.
Skill Claude CodeCodex
Use when a request needs cross-domain Cognia status, logs, sessions, or replayable events for read-only investigation; do not use when any write or side effect is requested, or when one domain skill fully covers the read.
Skill Claude CodeCodex
Use when a request needs safeguarded Git commits, rebases, resets, worktrees, destructive changes, conflict handling, or recovery through Cognia; do not use when handling non-Git tools or a simple read-only Git lookup.
Skill Claude CodeCodex
Use when a request stays within session discovery, timelines, message delivery, or character binding to operate routine Cognia chat data; do not use when the request spans domains or is primarily an agent, task, or connector workflow.
Skill Claude CodeCodex
Use when inspecting routine Cognia capabilities, admin leases, secrets, sync, logs, bridge administration, or device reports; do not use when performing multi-step backup export, restore, or disaster recovery, where cognia-host-backup-recovery applies.
Skill Claude CodeCodex
Use when a request concerns Cognia task lifecycle, workspaces, resources, comments, patches, transfers, conflict resolution, settlement, or cleanup; do not use when the work primarily controls agent runtimes or chat sessions rather than task state.
Skill Claude CodeCodex
Use when a task broadly concerns same-host Cognia operations or no narrower cognia-host- skill is known to discover and safely invoke the loopback-only Headless RPC plane; do not use when a domain or workflow skill directly matches the request.
Skill Codex
Use when the user wants to design, redesign, shape, critique, audit, polish, clarify, distill, harden, optimize, adapt, animate, colorize, extract, or otherwise improve a frontend interface. Covers websites, landing pages, dashboards, product UI, app shells, components, forms, settings, onboarding, and empty states.…
Skill Claude CodeCodex
Decompose multi-agent work, dispatch independent tasks within host limits, and synthesize one coherent result.
Skill Claude CodeCodex
Turn quantitative comparisons, trends, compositions, and correlations into live Cognia chart artifacts.
Skill Claude CodeCodex
Complete one authorized first-run folder, screenshot OCR, or web-summary request with a visible result.
Skill Claude CodeCodex
Drive a real computer through screenshot-grounded actions, verification, consent, and safe recovery.
Skill Claude CodeCodex
Create polished, responsive technical and product diagrams as sanitized HTML with inline SVG.
Skill Claude CodeCodex
Answer from injected Digital Twin knowledge while separating facts, inference, voice, and provenance.
Skill Claude CodeCodex
Advance an active goal or loop through honest judging, bounded continuation, completion, and blocking.
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: