CATLX — Runtime Lifecycle Sequences
This skill owns the ordered end-to-end sequences that tie the subsystems together: how CATLX boots, how a
voice command flows, how a workflow runs, how a plugin is installed, and how recovery proceeds.
Canonical detail: ../knowledge/references/runtime-lifecycle.md. The voice sequence is also in
../examples/voice-command-lifecycle.md. Load on demand.
Purpose
Give the exact, ordered step sequences so that no stage is skipped or performed out of order. Use these as the
authoritative procedural truth for orchestrating CATLX behavior.
When to activate
- User asks about startup/boot order, or "what happens when CATLX starts."
- Tracing a voice command, a workflow run, a plugin install, or a recovery end-to-end.
- Verifying a specific stage (e.g. when the Capability Router runs vs the Memory Broker).
What this skill handles
- Boot sequence (§18.1) — launcher detects
CATLX_ROOT → stale-PID check → Hardware Profiler (< 50 ms) →
Capability Router emits CapabilityMap → databases open (WAL replay) → Module Registry scanned in
dependency order → Event Bus on :7701 → Security Layer → Memory Broker → AI Provider Layer (health/local
servers T1+) → Workflow Engine scheduler → Voice Pipeline (wake word + mic) → Plugin Runtime →
Electron Shell/HUD → write PID lockfile → emit system.ready → restore workspace →
operational.
- Voice command lifecycle (§18.2) — wake word → capture → VAD (800 ms) → STT → context → NLU
(
IntentObject, confidence < 0.75 → disambiguate/loop) → Workflow Planner → Permission validation → DAG
execute → per-step checkpoint + TTS/HUD → completion → Memory Broker writes EpisodicRecord → monitoring.
- Workflow lifecycle (§18.3) — load YAML/dynamic → validate DAG (no cycles, deps present, modules
available) → create execution context (run ID, input snapshot, rollback log) → schedule → dispatch root
nodes → per-node module invoke → store output/unlock dependents → on failure retry policy → on FAILED offer
rollback → on success COMPLETED → emit completion → archive run.
- Plugin lifecycle (§18.4) — package (marketplace/local) → signature verify → manifest parse → capability
grant review → store grants in
plugins.db → resolve/vendor deps → build/validate → register → load into
sandbox → initialize() → register capabilities with router → available → update (hot-swap or restart
flag) → uninstall (cleanup(), deregister, destroy sandbox, remove files).
- Recovery lifecycle (§18.5) — crash detected via stale PID → Recovery Mode flag / Safe Mode UI → replay
WAL → load workspace snapshot → find
RUNNING workflows → verify checkpoint hash → present Resume/Rollback/
Ignore → execute chosen action → clear flag → continue boot.
Requirements / constraints
- R8 (replay safety): all lifecycle steps that mutate state are checkpointed.
- R7 (deterministic recovery): recovery sequence is deterministic;
Recovery Mode flag cleared only when
complete.
- Windows-first paths and mechanisms apply throughout (see
../knowledge/rules/windows-rules.md).
Canonical knowledge it reads
../knowledge/references/runtime-lifecycle.md · ../knowledge/references/recovery.md ·
../knowledge/references/workflow-engine.md · ../knowledge/references/voice-pipeline.md ·
../knowledge/references/plugin-ecosystem.md · ../knowledge/references/data-registries.md.
Delegation
Because these sequences cross subsystems, delegate to the owning skill when a specific stage needs detail:
catlx-voice-pipeline, catlx-workflow-engine, catlx-plugin-ecosystem, catlx-electron-shell,
catlx-security, catlx-recovery, catlx-hardware-adaptation, catlx-capability-routing. Use
skill({ name: "<target>" }) and pass the specific stage as the minimum context.
Edge cases & warnings
- Boot ordering: the Capability Router must emit the CapabilityMap before any subsystem initializes.
- Stale PID: a genuine crash triggers recovery; a live PID must not be treated as stale.
- Confidence loop: NLU must loop on disambiguation only while confidence < threshold; avoid infinite loops
(cap the number of clarification turns).
- Plugin/unit ordering at boot: modules and plugins load in dependency order.
Component lifecycle policy (reuse → install → adapt → create)
NEVER create a new component as the default. Before building/creating anything (a sub-skill, dependency,
reference, workflow, helper, adapter, or template), check, in order:
- Reuse an existing local component (resolve aliases/equivalent capabilities first) — reuse, don't rebuild.
- Use an already-registered component from the registry.
- Install a suitable existing, trusted, supported component → validate → register → connect to the graph → use.
- Adapt an existing compatible component via a small persistent adapter/wrapper instead of re-creating it.
- Create only as last resort — then make it permanent immediately: stable id, canonical location, register,
add to the capability index + dependency graph, add provenance, use, and allow future reuse.
- Never reorganise/recreate already-generated components (no
Skill X 2 / new / temp variants); extend the
existing one. Never create a second competing knowledge source; connect back to the canonical knowledge/ layer.
Promote any reusable artifact out of /tmp/scratch into the permanent ecosystem.
Full policy: ../knowledge/rules/component-lifecycle.md.
Source / provenance
- Source: PART XVIII §18.1–18.5 (boot, voice command lifecycle detailed, workflow lifecycle, plugin
lifecycle, recovery lifecycle).
- Inferred: none; sequences preserved as ordered steps.
1---2> Mirror copy of `skills\catlx-core\catlx-runtime-lifecycle\SKILL.md` (canonical content lives in `catlx-core`). Kept at the repository root only for registry-path compatibility. Edit the canonical copy, not this file.3name: catlx-runtime-lifecycle4description: "Handles the CATLX runtime lifecycle sequences: the full boot sequence, the detailed voice-command lifecycle, the workflow lifecycle, the plugin lifecycle, and the recovery lifecycle, each as ordered step sequences. Use when the user asks about CATLX startup/boot order, what happens on wake word, how a workflow runs step by step, how a plugin is installed/uninstalled, or the recovery sequence."5metadata:6 catlx: subsystem7 category: lifecycle8 subsystem: Runtime_Lifecycle9 capability: lifecycle-sequences10 version: "1.0.0"11 source: https://skillmd.com/skills/samraddhashrivastavatech/catlx-runtime-lifecycle12 upstream: "PART XVIII §18.1-18.5"13 aliases: "lifecycle, boot, startup, runtime sequences, initialize, boot sequence"14 depends-on: "catlx-recovery, catlx-voice-pipeline, catlx-workflow-engine, catlx-plugin-ecosystem, catlx-electron-shell, catlx-security"15---1617# CATLX — Runtime Lifecycle Sequences1819This skill owns the **ordered end-to-end sequences** that tie the subsystems together: how CATLX boots, how a20voice command flows, how a workflow runs, how a plugin is installed, and how recovery proceeds.2122> Canonical detail: `../knowledge/references/runtime-lifecycle.md`. The voice sequence is also in23> `../examples/voice-command-lifecycle.md`. Load on demand.2425---2627## Purpose2829Give the exact, ordered step sequences so that no stage is skipped or performed out of order. Use these as the30authoritative procedural truth for orchestrating CATLX behavior.3132## When to activate3334- User asks about startup/boot order, or "what happens when CATLX starts."35- Tracing a voice command, a workflow run, a plugin install, or a recovery end-to-end.36- Verifying a specific stage (e.g. when the Capability Router runs vs the Memory Broker).3738## What this skill handles39401. **Boot sequence** (§18.1) — launcher detects `CATLX_ROOT` → stale-PID check → Hardware Profiler (< 50 ms) →41 Capability Router emits `CapabilityMap` → databases open (WAL replay) → Module Registry scanned in42 dependency order → Event Bus on `:7701` → Security Layer → Memory Broker → AI Provider Layer (health/local43 servers T1+) → Workflow Engine scheduler → Voice Pipeline (wake word + mic) → Plugin Runtime →44 Electron Shell/HUD → write PID lockfile → emit `system.ready` → restore workspace →45 operational.462. **Voice command lifecycle** (§18.2) — wake word → capture → VAD (800 ms) → STT → context → NLU47 (`IntentObject`, confidence < 0.75 → disambiguate/loop) → Workflow Planner → Permission validation → DAG48 execute → per-step checkpoint + TTS/HUD → completion → Memory Broker writes EpisodicRecord → monitoring.493. **Workflow lifecycle** (§18.3) — load YAML/dynamic → validate DAG (no cycles, deps present, modules50 available) → create execution context (run ID, input snapshot, rollback log) → schedule → dispatch root51 nodes → per-node module invoke → store output/unlock dependents → on failure retry policy → on FAILED offer52 rollback → on success COMPLETED → emit completion → archive run.534. **Plugin lifecycle** (§18.4) — package (marketplace/local) → signature verify → manifest parse → capability54 grant review → store grants in `plugins.db` → resolve/vendor deps → build/validate → register → load into55 sandbox → `initialize()` → register capabilities with router → available → update (hot-swap or restart56 flag) → uninstall (`cleanup()`, deregister, destroy sandbox, remove files).575. **Recovery lifecycle** (§18.5) — crash detected via stale PID → Recovery Mode flag / Safe Mode UI → replay58 WAL → load workspace snapshot → find `RUNNING` workflows → verify checkpoint hash → present Resume/Rollback/59 Ignore → execute chosen action → clear flag → continue boot.6061## Requirements / constraints6263- **R8 (replay safety):** all lifecycle steps that mutate state are checkpointed.64- **R7 (deterministic recovery):** recovery sequence is deterministic; `Recovery Mode` flag cleared only when65 complete.66- Windows-first paths and mechanisms apply throughout (see `../knowledge/rules/windows-rules.md`).6768## Canonical knowledge it reads6970`../knowledge/references/runtime-lifecycle.md` · `../knowledge/references/recovery.md` ·71`../knowledge/references/workflow-engine.md` · `../knowledge/references/voice-pipeline.md` ·72`../knowledge/references/plugin-ecosystem.md` · `../knowledge/references/data-registries.md`.7374## Delegation7576Because these sequences cross subsystems, delegate to the owning skill when a specific stage needs detail:77`catlx-voice-pipeline`, `catlx-workflow-engine`, `catlx-plugin-ecosystem`, `catlx-electron-shell`,78`catlx-security`, `catlx-recovery`, `catlx-hardware-adaptation`, `catlx-capability-routing`. Use79`skill({ name: "<target>" })` and pass the specific stage as the minimum context.8081## Edge cases & warnings8283- **Boot ordering:** the Capability Router must emit the CapabilityMap before any subsystem initializes.84- **Stale PID:** a genuine crash triggers recovery; a live PID must not be treated as stale.85- **Confidence loop:** NLU must loop on disambiguation only while confidence < threshold; avoid infinite loops86 (cap the number of clarification turns).87- **Plugin/unit ordering at boot:** modules and plugins load in dependency order.8889## Component lifecycle policy (reuse → install → adapt → create)9091**NEVER create a new component as the default.** Before building/creating anything (a sub-skill, dependency,92reference, workflow, helper, adapter, or template), check, in order:931. **Reuse** an existing local component (resolve aliases/equivalent capabilities first) — reuse, don't rebuild.942. **Use** an already-registered component from the registry.953. **Install** a suitable existing, trusted, supported component → validate → register → connect to the graph → use.964. **Adapt** an existing compatible component via a small persistent adapter/wrapper instead of re-creating it.975. **Create only as last resort** — then make it permanent immediately: stable id, canonical location, register,98 add to the capability index + dependency graph, add provenance, use, and allow future reuse.996. Never reorganise/recreate already-generated components (no `Skill X 2` / `new` / `temp` variants); extend the100 existing one. Never create a second competing knowledge source; connect back to the canonical `knowledge/` layer.101 Promote any reusable artifact out of `/tmp`/scratch into the permanent ecosystem.102103> Full policy: `../knowledge/rules/component-lifecycle.md`.104105## Source / provenance106107- **Source:** PART XVIII §18.1–18.5 (boot, voice command lifecycle detailed, workflow lifecycle, plugin108 lifecycle, recovery lifecycle).109- **Inferred:** none; sequences preserved as ordered steps.