/arch — System Architect (the map and health of the whole system)

One place that knows what the system is made of, what is healthy, what fell over — and tests it. Deterministic, 0 tokens, READ-ONLY. This is RECALL for the infrastructure layer: look BEFORE changing shared infrastructure, rescan AFTER.

Engine: $IMPORTS_ROOT/arch/ (git-backed in _imports). The source of truth is system.db (built by the nightly scan at 05:45). Dashboard _Dashboards\System-Health.html, MOC 00-System\_System-MOC.md, auto-inventory 00-System\System-Automations.md (replaces the manual [[automation-inventory]]).

Commands

# status (default) — health summary + score
python "$IMPORTS_ROOT/arch/arch_status.py"

# only what is broken / red
python "$IMPORTS_ROOT/arch/arch_status.py" broken

# dead code — scripts wired to nothing
python "$IMPORTS_ROOT/arch/arch_status.py" dead

• /arch → arch_status.py (status).
• /arch broken → arch_status.py broken.
• /arch dead → arch_status.py dead.
• /arch scan (force a fresh scan, AFTER an infrastructure change) → rebuild the catalog:

  cd /d $IMPORTS_ROOT/arch
  python sys_scan.py && python sys_coverage.py && python build_system_docs.py && python build_arch_map.py && python sys_check.py

  (That is the run_architect.cmd pipeline minus the final vault_backup.py. The full nightly run = run_architect.cmd, scheduled task "System Architect Nightly" at 05:45.)

When to use it (STANDING — an always-loaded rule)

• BEFORE adding/removing/changing shared infrastructure (a scheduled task, an _imports script, an MCP, a DB, a hook, a skill, a pipeline) → /arch / /arch broken: what exists and what depends on it — will I break a neighbour?
• AFTER the change → /arch scan, so the map does not fall behind reality.
• RED → deal with the red first; result!=0 on a task does not always mean "broken" (a benign lock collision happens — check the source). Never delete an active/critical asset before understanding its dependencies.

Pitfalls (from memory)

• A coverage metric must MEASURE the artifact, not hardcode a verdict (the false backup-line alarm, 2026-06-22).
• RED.flag is written only on critical/daily failures (SRE style: alert on the symptom). The phase-5 routine system-architect-red-alert (06:21) pings Telegram Saved Messages when red and stays silent when green.

Canon

Memory [[system-architect]] · decision decision-architect-system-platform · the Bible rule "check the map before changing the system" (for human assistants and LLMs alike). Related: [[verify-existing-before-proposing]], [[automation-inventory]], [[vault-data-architecture]].

Like this skill? It is one of 100 in second-brain-starter-kit: the second brain we built for ourselves and run every day at Palo Alto AI Research Lab. Install the whole set with npx skills add tonydzi/second-brain-starter-kit. Everything is open source and free, so take what you need.

Flagships worth a look on their own: secondop-panel (a second opinion from a panel of external models), claude-memory-tidy (stop your agent's memory from rotting), telegram-mcp-kit (your own Telegram over MCP in about 15 minutes).

Author: Anton Dziatkovskii, Palo Alto AI Research Lab. Telegram @tonydzi - WhatsApp +1 341 222 9178 - X @Tony_Stef_

Engineers: want to test-drive this setup? Message me. I hand out free starter seeds to engineers who test and report back, and custom skill requests are welcome.