Raxol symphony

Skill DROOdotFOO/agent-skills/skills/raxol-symphony

Agent skills, autonomous agents, and MCP-companions for programming

Install

npx -y skills add DROOdotFOO/agent-skills --skill raxol-symphony

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Symphony: the Raxol tracker-driven coding-agent orchestrator (raxol_symphony, pre-alpha). Turns tracker issues into autonomous agent runs in isolated workspaces, surfaces evidence (CI/PR/asciinema), and supports human-in-the-loop paused runs across six surfaces. TRIGGER when: mix.exs lists :raxol_symphony; code imports Raxol.Symphony.*; there is a WORKFLOW.md driving an orchestrator; user asks about tracker-driven agent orchestration, Symphony runners (RaxolAgent/Codex), evidence capture, or paused-run resume with Raxol. DO NOT TRIGGER when: building a single Raxol agent or the core framework (use raxol skill); agent payments / ACP job sessions (use raxol-payments skill); generic multi-agent system design without Symphony (use agent-designer skill); the Claude Agent SDK (use claude-api).

SKILL.md

3.9 KB, 750 tokens by cl100k_base, as published. Nobody here has run it

Raxol Symphony Skill

raxol_symphony (v0.1, pre-alpha) is an OTP port of OpenAI's Symphony: an orchestrator that turns tracker work into autonomous coding-agent runs. Each issue gets an isolated workspace, runs an agent until a workflow-defined handoff state, and surfaces evidence (CI, PR comments, asciinema walkthrough) so engineers manage outcomes, not prompts.

Config is a WORKFLOW.md (YAML front-matter + a Liquid prompt template), hot-reloaded on change. Pre-alpha: APIs may shift.

What You Get

Orchestrator loop: poll tracker -> isolate workspace -> run runner -> collect evidence
Runners: RaxolAgent (primary), Codex (JSON-RPC parity), AgentSession, Review, Noop
Trackers: Linear (GraphQL), GitHub (state labels), Memory (tests)
WORKFLOW.md hot-reload + a workflow graph adapter (saga-style)
Evidence framework: CI, PR comments, complexity (cloc), asciinema recordings
Six surfaces + a first-class paused-run substrate (pause/resume across surfaces)

Quickstart

{:ok, _pid} = Raxol.Symphony.start_link(workflow_path: "WORKFLOW.md")

<!-- WORKFLOW.md -->
---
tracker:
  kind: linear
  project_slug: demo
  api_key: $LINEAR_API_KEY
  active_states: ["Todo", "In Progress"]
  terminal_states: ["Done", "Cancelled"]
runner:
  kind: raxol_agent
  agent:
    backend: anthropic
    model: claude-sonnet-4-6
    max_turns: 20
---
You are working on issue {{ issue.identifier }}: {{ issue.title }}.

Reading Guide

Task	File
Orchestrator, trackers, runners, config	`orchestrator.md`
WORKFLOW.md, graph adapter, evidence	`workflows-evidence.md`
Surfaces + pause/resume + sandboxes	`surfaces-pause.md`

Key Conventions

One workspace per issue; PathSafety sanitizes identifiers and asserts paths stay inside the workspace root. Never run an agent outside its workspace.
A runner returns :ok (retry-continue), {:error, reason} (backoff retry), or {:pause, reason, token} (park in the orchestrator's paused map).
The Review runner's Contract deliberately carries no workspace reference -- reviewer isolation is an invariant.

Common Pitfalls

Blocking on human input inline -- return {:pause, reason, token} and let a surface/Resumer resume; don't sleep the runner.
Bad WORKFLOW.md -- WorkflowStore falls back to the last-known-good config and records last_error/1; check it after edits.
Assuming production-ready -- v0.1 pre-alpha; pin versions and expect churn.

Gives 0 of the 12 instructions most agent orchestration skills give in 750 tokens

Counted across 742 of the 995 authors here whose files we hold, read 2026-08-07

reference existing artifacts by path or URLin 53 of 742, across 25 files
run the full test suite after integrating changesin 51 of 742, across 19 files
dispatch one agent per independent problem domainin 50 of 742, across 17 files
verify fixes do not conflictin 45 of 742, across 13 files
include a suggested skills section in the documentin 45 of 742, across 17 files
redact sensitive informationin 41 of 742, across 11 files
save to the temporary directory of the operating systemin 39 of 742, across 10 files
tailor the document to user-provided focus argumentsin 39 of 742, across 9 files
spot check agent changes for systematic errorsin 34 of 742, across 7 files
write a handoff document summarising the current conversationin 31 of 742, across 6 files
Assign each agent a specific scopein 23 of 742, across 8 files
provide specific scope and clear goalin 23 of 742, across 5 files

Said here and by no other author read

start the orchestrator with a workflow path
sanitize workspace identifiers
keep paths inside the workspace root
run agents only inside their workspace
return a pause tuple for human input
do not sleep the runner

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.