Execute qualquer Skill no Manus
com um clique

Execute qualquer Skill no Manus com um clique

cortex

Estrelas1

Forks2

Atualizado16 de junho de 2026 às 21:15

This skill should be used when the user asks to "search logs", "check errors", "tail logs", "show recent logs", "find log entries", "correlate events", "list hosts", "log stats", "syslog", "check homelab logs", or mentions system logs, syslog, log analysis, or log intelligence across homelab hosts.

Instalação

Instalar com Codex ou Claude Copie este prompt, cole no Codex, Claude ou outro assistente e deixe que ele revise a página da skill e instale para você.

Executar no Manus

Fonte

jmagar

jmagar/cortex

Abrir repositório GitHub Ver repositórios do creator

Download

Executar no Manus

Ocupações relacionadasSOC

Baseado na classificação ocupacional SOC

Administradores de redes e sistemas de computadorInformática e Matemática·SOC 15-1244

Explorador de arquivos

2 arquivos

SKILL.md

readonly

Mais deste repositório

mesmo repositório

cortex-report

jmagar/cortex

This skill should be used when the user asks for a homelab health report, syslog summary, fleet status report, log analysis summary, 'what happened in the last 24 hours', 'show me this week's errors', 'summarize recent activity', or any time-bounded log analysis that should produce a written markdown report.

2026-06-161

cortex-redeploy

jmagar/cortex

Re-run the cortex plugin setup hook with the current userConfig and verify the Docker Compose deployment. Use when the user asks to redeploy cortex, apply plugin config changes immediately, rerun the setup hook, refresh the Docker deployment, or recover after an automated SessionStart/ConfigChange hook did not run.

2026-06-101

cortex-deploy-dropins

jmagar/cortex

Deploy rsyslog forwarding drop-ins to configured fleet hosts over SSH. Use when configuring fleet forwarding, repairing missing rsyslog forwarding, or updating forwarding after server_url or syslog port changes.

2026-06-041

cortex-dr

jmagar/cortex

Run a comprehensive cortex health check covering environment, config quality, storage, ports, service status, HTTP health, MCP actions, listener reachability, Docker ingest, and fleet rsyslog forwarding. Use when the user asks for syslog doctor, deployment diagnostics, first-run preflight, health check, sanity check, or broad deployment verification.

2026-06-041

cortex-frustration-assessment

jmagar/cortex

This skill should be used after running cortex action=abuse_investigate to analyze the resulting evidence bundle. Use when the user asks to assess frustration incidents, evaluate abuse signals, analyze agent or user friction, produce a frustration report, or follow up on abuse_investigate results.

2026-06-041

cortex-logs

jmagar/cortex

Tail or follow cortex service logs from Docker Compose. Use when the user asks for cortex service logs, startup logs, crash logs, plugin deployment logs, Docker logs, or follow mode. This is for the service's stdout/stderr, not client syslog entries.

2026-06-041

name	cortex
description	This skill should be used when the user asks to "search logs", "check errors", "tail logs", "show recent logs", "find log entries", "correlate events", "list hosts", "log stats", "syslog", "check homelab logs", or mentions system logs, syslog, log analysis, or log intelligence across homelab hosts.

Cortex Skill

Rust-based syslog receiver and MCP server for homelab log intelligence. Receives RFC 3164/5424 syslog from all homelab hosts, stores in SQLite with FTS5 full-text search, and exposes one MCP tool with action dispatch for AI-driven log analysis.

Tool

A single MCP tool, mcp__cortex__cortex, dispatches on a required action argument:

action	purpose
`search`	Full-text search with FTS5
`filter`	Structured filter-only log retrieval
`tail`	Most recent entries
`errors`	Error/warning summary by host and severity
`hosts`	List all known hosts with first/last seen
`map`	Homelab host/source/app/heartbeat map
`host_state`	Latest bounded heartbeat state for one host
`fleet_state`	Fleet-wide heartbeat snapshot with pressure flags and summary counts
`correlate_state`	Correlate logs with heartbeat summaries around a reference time
`sessions`	AI transcript sessions by project
`search_sessions`	Ranked grouped session search
`abuse`	Abuse hits in AI transcripts with same-session context
`abuse_incidents`	Groups abuse hits into scored incident candidates
`abuse_investigate`	Expands incidents into deterministic evidence bundles
`ai_correlate`	AI transcript anchors cross-referenced against non-AI logs
`usage_blocks`	AI activity in 5-hour windows
`project_context`	Summary for one AI project path
`list_ai_tools`	Distinct AI tools with counts
`list_ai_projects`	Distinct AI projects with counts
`correlate`	Cross-host event correlation in a time window
`stats`	Database statistics
`status`	Lightweight runtime and DB health
`apps`	Distinct app names with log and host counts
`source_ips`	Source identifiers with hostname breakdown
`timeline`	Bucketed counts over time
`patterns`	Near-duplicate message template clusters
`context`	Surrounding logs around an event
`get`	One log by id, including raw frame
`ingest_rate`	Recent ingest throughput
`silent_hosts`	Hosts quiet beyond a threshold
`clock_skew`	Per-host clock skew distribution
`anomalies`	Recent vs baseline volume/error comparison
`compare`	Compare two time ranges
`compose_status`	Redacted read-only Compose deployment diagnostics
`compose_doctor`	Strict Compose deployment health diagnostics
`unaddressed_errors`	List unacknowledged repeating error signatures
`ack_error`	Acknowledge an error signature
`unack_error`	Revoke an existing acknowledgement
`file_tails`	Manage Cortex-owned file-tail ingest sources
`notifications_recent`	List recent notification firings
`notifications_test`	Send a test notification via Apprise
`similar_incidents`	FTS5 cluster search over historical system logs matching a query
`ask_history`	Search AI transcript history for sessions related to a topic
`incident_context`	Full log context bundle for a known time window
`graph`	Resolve graph entities and return bounded one-hop neighborhoods with evidence
`help`	Canonical in-tree action reference (use as ground truth if this doc drifts)

Always prefer the MCP tool. Fall back to HTTP only when MCP is unavailable.

The skill works identically in both server mode (this machine hosts the receiver) and client mode (connects to a remote server) — the connection details are configured at plugin install time. Source identity is captured per log: syslog entries carry the verified network sender as IP:port; Docker socket-proxy ingested entries carry docker://host/container/stream.

Action Reference

For parameter-level details and response shapes, use the live action reference:

mcp__cortex__cortex(action="help")

When working from the repository instead of a live server, use docs/mcp/TOOLS.md, docs/mcp/SCHEMA.md, and docs/mcp/CORRELATION.md as the canonical references. Keep this skill focused on when to use the tool, safe invocation patterns, and common workflows rather than duplicating every action schema.

FTS5 reminders for search and other query-bearing actions:

AND, OR, NOT are uppercase boolean operators.
Quote phrases and hyphenated terms, for example "smoke-test".
Invalid FTS5 syntax returns a database error.

HTTP Fallback Mode

Use only when the MCP tool is unavailable. The plugin exports connection settings to Bash subprocesses as:

CLAUDE_PLUGIN_OPTION_SERVER_URL — base URL (e.g. http://localhost:3100)
CLAUDE_PLUGIN_OPTION_API_TOKEN — bearer token

Sensitive value handling: api_token is declared sensitive: true in the plugin manifest. It is never substituted into skill content as ${user_config.api_token} — only the env var path above is valid. Do not inline the token in this document or any skill text.

Health check (no auth required)

# /health is unauthenticated — no Authorization header needed
curl -s "$CLAUDE_PLUGIN_OPTION_SERVER_URL/health"

Tail logs

curl -s -X POST "$CLAUDE_PLUGIN_OPTION_SERVER_URL/mcp" \
  -H "Authorization: Bearer $CLAUDE_PLUGIN_OPTION_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"cortex","arguments":{"action":"tail","n":20}}}'

Search logs

curl -s -X POST "$CLAUDE_PLUGIN_OPTION_SERVER_URL/mcp" \
  -H "Authorization: Bearer $CLAUDE_PLUGIN_OPTION_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call","params":{"name":"cortex","arguments":{"action":"search","query":"error","limit":20}}}'

Get stats

curl -s -X POST "$CLAUDE_PLUGIN_OPTION_SERVER_URL/mcp" \
  -H "Authorization: Bearer $CLAUDE_PLUGIN_OPTION_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"cortex","arguments":{"action":"stats"}}}'

Severity Levels

Level	Numeric	Description
emerg	0	System unusable
alert	1	Immediate action required
crit	2	Critical condition
err	3	Error condition
warning	4	Warning condition
notice	5	Normal but significant
info	6	Informational
debug	7	Debug messages

errors returns only emerg through warning. correlate defaults severity_min to warning (returns warning through emerg).

Log Intelligence Workflows

Quick homelab health check

mcp__cortex__cortex(action="errors")
mcp__cortex__cortex(action="tail", host="unraid", n=50)
mcp__cortex__cortex(action="search", query='OOM OR "out of memory"', host="unraid")

Incident investigation

# 1. Find the incident
mcp__cortex__cortex(action="search", query='panic OR crash OR "segmentation fault"', limit=10)

# 2. Correlate across hosts at that timestamp
mcp__cortex__cortex(action="correlate",
                    reference_time="<timestamp from step 1>",
                    window_minutes=15,
                    severity_min="warning")

# 3. Confirm which hosts were active
mcp__cortex__cortex(action="hosts")

Trace a specific Docker container's logs

# Docker ingest sets source_ip to docker://host/container/stream
mcp__cortex__cortex(action="search", source="docker://squirts/postgres/stdout", limit=50)

Storage health

mcp__cortex__cortex(action="stats")
# Check: write_blocked, logical_db_size_mb vs threshold, free_disk_mb vs threshold

Ports

Port	Protocol	Purpose
1514	UDP + TCP	Syslog receiver
3100	TCP	MCP HTTP endpoint (`POST /mcp`, `GET /health`)

Port 1514 (not 514) avoids needing CAP_NET_BIND_SERVICE. iptables PREROUTING redirects 514→1514 for devices that can't be reconfigured — that's deployment-time setup, not runtime.