autosymph-monitor

Autosymph Monitor Agent

Observe. Diagnose. Heal. This is the operator cockpit for autosymph.

You attach to a running autosymph instance, poll for status, detect bad loops, fix them (structify, investigate), and put issues back into the pipeline.

---

Step 0: Load Config

Read the user's autosymph configuration to understand what you're monitoring.

Find configs

# Project configs
ls "${AUTOSYMPH_CONFIG_DIR:-$HOME/.autosymph/config}"/projects/*.yaml 2>/dev/null

# Device config
ls "${AUTOSYMPH_CONFIG_DIR:-$HOME/.autosymph/config}"/devices/*.yaml 2>/dev/null

# Legacy single-file config
ls "${AUTOSYMPH_CONFIG_DIR:-$HOME/.autosymph/config}"/*.yaml 2>/dev/null | grep -v _template

Parse and remember

For each project config, extract:

• Project name (tracker.project)
• State definitions (states.{name}.type, .prompt, .model, .max_turns, .allowed_tools)
• Transitions (states.{name}.transitions)
• Polling interval (polling.interval_ms)
• Stall timeout (claude.stall_timeout_ms)
• Max concurrent agents (agent.max_concurrent_agents)
• Log root (logging.log_root, default ~/.autosymph/logs)

From device config:

• Simulators (resources.ios_simulator)
• Dev ports (resources.dev_port_range)

Announce what you found:

Loaded 3 project configs: ios-app, web-app, api-service
  ios-app: verify=opus, implement=sonnet, max_agents=3, stall=300s
  web-app: verify=opus, implement=sonnet, max_agents=3
  api-service: verify=opus, implement=sonnet, max_agents=3

---

Step 1: Connect to Autosymph

STATUS_URL="$(${AUTOSYMPH_SKILLS_DIR:-$HOME/.autosymph/skills}/autosymph-monitor/scripts/status-url.sh)"
curl -s "$STATUS_URL/status"

If this fails:

• Check if autosymph process is running: ps aux | grep autosymph
• If running but no API → autosymph was started before server.py was installed. Tell user: "autosymph is running but the status API isn't available. Restart autosymph (q in TUI, then uv run autosymph start) to enable the API."
• If not running → "autosymph is not running. Start it with uv run autosymph start."