Trace, replay, and cost-attribute multi-agent runs: hook-driven JSONL AgentTrace, deterministic OrchestrationReplay, and per-subagent TokenLens — pure artifacts, with an optional read-only MCP dashboard server.
Plugin apex-agent-observability from marketplace apex-dev-skills in Skobyn/Apex-Dev-Skills. Commands read from the repo's manifest — 16 days ago.
This plugin runs code on your machine — its hooks fire automatically on events like tool use and it starts local MCP servers. The exact commands are listed below; read them before installing.
agent-traceRead and interpret an AgentTrace timeline to reconstruct WHERE a multi-agent or subagent run went wrong. Use when a swarm / subagent orchestration produced a bad or confusing result and you need to see the execution-order edges (which agent spawned which, in what order) and the per-agent tool sequen
orchestration-replayReconstruct the decomposition and tool sequence of a captured multi-agent run so a failed orchestration can be analyzed and re-run deterministically. Use after a swarm / subagent run failed and you want to replay exactly what happened — the spawn tree, each subagent's ordered tool calls, and the poi
token-lensAggregate per-subagent token and latency attribution from a captured multi-agent trace into a cost report. Use when you need to know WHICH subagent or branch of a swarm burned the token budget or the wall-clock time, when a multi-agent run felt expensive or slow, or when comparing the cost of two or
/apex-agent-observability:trace-reportProduce a full observability report for a captured multi-agent run — spawn tree, per-subagent tool sequence, and token/latency cost attribution. Pass an optional trace path as $ARGUMENTS (defaults to the most recent trace).
apex-agent-observabilityruns: python3 ${CLAUDE_PLUGIN_ROOT}/mcp/server.py
bash "${CLAUDE_PLUGIN_ROOT}/hooks/trace-event.sh" SubagentStart
bash "${CLAUDE_PLUGIN_ROOT}/hooks/trace-event.sh" SubagentStop
bash "${CLAUDE_PLUGIN_ROOT}/hooks/trace-event.sh" PreToolUse
bash "${CLAUDE_PLUGIN_ROOT}/hooks/trace-event.sh" PostToolUse
Route and govern delegation: a compiled routing policy picks the tier, provider, fan-out and review shape for each task or ad-hoc ask; PreToolUse/SubagentStop hooks enforce it; provider shims run external workers; every route, spawn, worker run and verdict lands in a hash-chained ledger, which report.sh --per-task prices per routed task. Drives apex-scope-loop as its execution engine.
2 skills · 6 commands · 10 agents · hooks
Typed, probability-carrying answers for the questions apex-dispatch and apex-scope-loop leave to judgment (task class, risk tier): one CLI, rubric files, a fail-closed validator for every backend, tri-state uncertainty, and calibration that only a measurement job can grant. Off until a repo opts in; absent, slow or unconfigured, routing stays table-only.
1 skill · 5 commands · hooks
Routing and gate engine for apex-app: lanes + surface-ledger routing, cross-platform guardrails, and phases that close on computed obligations.
2 skills · 4 commands · hooks
Decide -> plan -> loop: SPARC-shaped ADR + phased plan authoring with scope sizing (bugfix, poc, feature) and an intent round, autonomous /loop + /schedule execution, per-phase swarm dispatch, The Gibson's harness gates (baseline green gate, independent review, Tier-C human approval), and an operate phase that watches landed plans against their declared checks.
2 skills · 3 commands · 2 agents
Interview-first greenfield project initializer: opens with 'what are you trying to build?', researches the context, then scaffolds a fully-governed repo (paved-road defaults), provisions the Apex-Dev-Skills suite (plus optional ruflo), and verifies it before the first commit. Generated projects are steered to build features via apex-plan/apex-loop. Supports TypeScript/JavaScript, Python, Go, and Rust.
3 skills · 1 command · 1 agent
Tool-contract validation + runtime reliability: a PostToolUse ledger that infers per-tool schemas and flags contract drift, plus FlakeGuard detection of non-deterministic agent failures (retries, partial failures, flaky tools).
2 skills · 1 command · hooks
Core skills library for Claude Code: TDD, debugging, collaboration patterns, and proven techniques
15 skills · hooks
CMEM Pro for the Claude app - hooks stream tool use to cmem.ai and inject observations into new sessions and agents
3 skills · 1 MCP server · hooks
Up-to-date documentation lookup. Pull version-specific documentation and code examples directly from source repositories into your coding agent's context.
1 MCP server
Build powerful, stateful CLI interfaces for any GUI application using the cli-anything harness methodology.
5 commands
Automates browser interactions for web testing, form filling, screenshots, and data extraction
12 skills
Use Alibaba Open Code Review from local Codex.
2 skills
This page updates itself from your repo. Add the badge to your README, or get it Featured to pin it above the organic results with a followed link to your site.