Run any Skill in Manus with one click

talos-cleanup-orchestrator

Stars3

Forks1

UpdatedApril 25, 2026 at 03:44

Orchestrate a multi-agent parallel cleanup pass. This agent coordinates 7 specialist lanes, manages a shared communication channel, and ensures disjoint execution. Use when the user wants to parallelize the 7-track code quality cleanup.

Installation

Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.

Run Skill in Manus

Source

talosprotocol

talosprotocol/talos

View GitHub Repository View Creator Repositories

Download

Run Skill in Manus

Related occupationsSOC

Based on SOC occupation classification

Software DevelopersComputer and Mathematical Occupations·SOC 15-1252

File Explorer

3 files

SKILL.md

readonly

name	talos-cleanup-orchestrator
description	Orchestrate a multi-agent parallel cleanup pass. This agent coordinates 7 specialist lanes, manages a shared communication channel, and ensures disjoint execution. Use when the user wants to parallelize the 7-track code quality cleanup.

Talos Cleanup Orchestrator

This agent manages the "Cleanup Task Force" by splitting the 7-track cleanup into parallel lanes and maintaining a shared communication state.

Core Responsibilities

Plan: Analyze the codebase and split the 7 tracks into safe parallel lanes using talos-parallelize.
Coordinate: Use scripts/agent_comm.py to maintain a shared state file (CLEANUP_STATE.json).
Communicate: Post messages and updates to the shared channel so lane agents can "see" each other's progress and warnings.
Handoff: Generate context-rich prompts for the specialist agents (Deduplication, Dead Code, etc.) including the current shared state.

Shared Communication Channel

The orchestrator maintains CLEANUP_STATE.json with:

Tasks: Current status of each cleanup track.
Messages: Cross-agent warnings (e.g., "Deduplication found a conflict in src/core").
Shared Resources: Files currently being modified to prevent collisions.

Workflow

Initialize: Run python scripts/agent_comm.py init CLEANUP_STATE.json.
Parallelize: Map the 7 tracks to lanes.
- Track 3 (Dead Code) and Track 1 (Deduplication) often overlap.
- Track 5 (Type Strengthening) can run in parallel with Track 6 (Error Handling).
Dispatch: For each active lane, generate a prompt that includes:
- The specialist skill to use.
- The current CLEANUP_STATE.json content.
- Specific write boundaries.
Monitor: Periodically check CLEANUP_STATE.json for updates from agents and rebalance lanes if conflicts arise.

Tools

python scripts/agent_comm.py post <sender> <message>: Post a message to the channel.
python scripts/agent_comm.py update <task_id> <status> [details]: Update task progress.
python scripts/agent_comm.py get: Read the full state.

Guardrails

NEVER allow two agents to write to the same file simultaneously.
ENSURE all agents report their "Start" and "Finish" to the state file.
RECONCILE all findings in a final Integration Pass.

More from this repository

same repository

talos-ai-engineer-agent

talosprotocol/talos

Act as the Talos AI engineer for LLM integrations, agent tooling, prompt artifacts, tool allowlists, evals, redaction, and schema-validated outputs. Use when the user explicitly wants a Talos specialist for model-driven flows that must remain deterministic, auditable, and constrained.

2026-04-253

talos-api-tester-agent

talosprotocol/talos

Act as the Talos API tester for endpoint, auth, idempotency, pagination, proxy, and integration validation work. Use when the user explicitly wants a testing specialist who prioritizes negative cases, real enforcement paths, and security-sensitive regressions over shallow mocks.

2026-04-253

talos-artifact-janitor-agent

talosprotocol/talos

Act as the Talos artifact janitor for dirty-worktree classification, generated-artifact cleanup planning, and ignore hygiene. Use when the user explicitly wants a specialist to separate intentional edits from generated noise and recommend the safest cleanup path without reverting others' work.

2026-04-253

talos-backend-architect-agent

talosprotocol/talos

Act as the Talos backend architect for API, schema, storage, and service-boundary work. Use when the user explicitly wants backend design or implementation from a Talos specialist who follows contract-first, audit, and failure-mode rigor. Do not use for frontend-only or marketing tasks.

2026-04-253

talos-capability-audit

talosprotocol/talos

Use when changing capability issuance, authz enforcement, gateway policy checks, audit logging, agent identity, session validation, revocation, or security-sensitive tests in Talos. Preserve signing and verification invariants, require negative cases, and confirm audit coverage and strict schema behavior.

2026-04-253

talos-ci-failure-manager-agent

talosprotocol/talos

Act as the Talos CI failure manager for build, lint, typecheck, contract drift, and test triage work. Use when the user explicitly wants a CI/build triage specialist who isolates the first failing signal, chooses the smallest local repro, and recommends the safest next fix without masking flakes or infra problems.

2026-04-253

name	talos-cleanup-orchestrator
description	Orchestrate a multi-agent parallel cleanup pass. This agent coordinates 7 specialist lanes, manages a shared communication channel, and ensures disjoint execution. Use when the user wants to parallelize the 7-track code quality cleanup.

Talos Cleanup Orchestrator

This agent manages the "Cleanup Task Force" by splitting the 7-track cleanup into parallel lanes and maintaining a shared communication state.

Core Responsibilities

Plan: Analyze the codebase and split the 7 tracks into safe parallel lanes using talos-parallelize.
Coordinate: Use scripts/agent_comm.py to maintain a shared state file (CLEANUP_STATE.json).
Communicate: Post messages and updates to the shared channel so lane agents can "see" each other's progress and warnings.
Handoff: Generate context-rich prompts for the specialist agents (Deduplication, Dead Code, etc.) including the current shared state.

Shared Communication Channel

The orchestrator maintains CLEANUP_STATE.json with:

Tasks: Current status of each cleanup track.
Messages: Cross-agent warnings (e.g., "Deduplication found a conflict in src/core").
Shared Resources: Files currently being modified to prevent collisions.

Workflow

Initialize: Run python scripts/agent_comm.py init CLEANUP_STATE.json.
Parallelize: Map the 7 tracks to lanes.
- Track 3 (Dead Code) and Track 1 (Deduplication) often overlap.
- Track 5 (Type Strengthening) can run in parallel with Track 6 (Error Handling).
Dispatch: For each active lane, generate a prompt that includes:
- The specialist skill to use.
- The current CLEANUP_STATE.json content.
- Specific write boundaries.
Monitor: Periodically check CLEANUP_STATE.json for updates from agents and rebalance lanes if conflicts arise.

Tools

python scripts/agent_comm.py post <sender> <message>: Post a message to the channel.
python scripts/agent_comm.py update <task_id> <status> [details]: Update task progress.
python scripts/agent_comm.py get: Read the full state.

Guardrails

NEVER allow two agents to write to the same file simultaneously.
ENSURE all agents report their "Start" and "Finish" to the state file.
RECONCILE all findings in a final Integration Pass.