OPS on-demand: This skill should be used when the user asks to "desk sweep", "open decisions", or…

MITAuto-check: notesProductivity & Automation

Install Ops Desk

skills CLI
$ npx skills add Lifecycle-Innovations-Limited/claude-ops --skill ops-desk -a claude-code

Project install by default; add -g for ~/.claude/skills/.

GitHub CLI
$ gh skill install Lifecycle-Innovations-Limited/claude-ops ops-desk --agent claude-code

Project scope by default; add --scope user for a personal install. Needs GitHub CLI 2.90.0 or later (public preview).

Manual copy
$ git clone --depth 1 https://github.com/Lifecycle-Innovations-Limited/claude-ops.git skills-src && mkdir -p .claude/skills && cp -r skills-src/claude-ops/skills/ops-desk .claude/skills/ops-desk && rm -rf skills-src

Use ~/.claude/skills/ instead of .claude/skills for a personal install. The folder must contain SKILL.md.

Claude Code skills documentation · loads skills from .claude/skills/

Facts

Skill name
ops-desk
GitHub stars
542
Token cost
~2.9k tokens
SKILL.md length
932 words
Files
1
Skills in repo
67
Repo updated
First seen
Licence
MIT

At a glance

OPS on-demand: This skill should be used when the user asks to "desk sweep", "open decisions", or…

  • Works in 4 steps: Build the item list (unless passed… → Run the gather workflow (READ-ONLY,… → Work the approval queue (main session,… → …
  • Asks to desk sweep
  • SKILL.md covers When to use ops-desk vs…, Step 1 — Build the item list…, Step 2 — Run the gather… and Step 3 — Work the approval…, plus 2 more sections
  • Calls gh

What it does

Ops Desk is an agent skill from Lifecycle-Innovations-Limited/claude-ops. OPS on-demand: This skill should be used when the user asks to "desk sweep", "open decisions", or…

Its SKILL.md is about 2.9k tokens, which your agent loads only when the skill is triggered. It is a single SKILL.md file with no bundled scripts.

It sits in Productivity & Automation, covering Email management. The repository describes itself as: Business operating system for Claude Code — 57 skills, 21 agents, smart daemon. Unified inbox (WhatsApp/Email/Slack/Telegram), autonomous PR merge, full-AWS monitoring, revenue… The licence is MIT.

When your agent uses it

  • Asks to desk sweep
  • Tasks that involve Email management

Example prompts

  • “desk sweep”
  • “open decisions”
  • “/ops-desk”

Requirements

  • Pre-approved tools (allowed-tools): Bash, Read, Grep, Glob, Skill, Agent, Workflow, AskUserQuestion, TeamCreate, SendMessage, TaskCreate, TaskUpdate, TaskList, CronCreate, mcp__gog__gmail_search, mcp__gog__gmail_read_thread, mcp__whatsapp__list_messages, mcp__whatsapp__archive_chat, mcp__linear__list_issues, mcp__linear__get_issue, mcp__linear__update_issue

Workflow steps

4 steps, taken from the step headings in SKILL.md.

  1. Build the item list (unless passed explicitly in )
  2. Run the gather workflow (READ-ONLY, batched)
  3. Work the approval queue (main session, outbound gate)
  4. Wrap up

What it can do on your machine

Read from SKILL.md and the folder at commit ee9c784. It shows what the files ask for, not the result of running them.

  • Tool permissions

    Pre-approves these tools, so the agent can use them without asking each time:

    • Bash
    • Read
    • Grep
    • Glob
    • Skill
    • Agent
    • Workflow
    • AskUserQuestion
    • TeamCreate
    • SendMessage

    …and 11 more on the same allowed-tools line.

    From allowed-tools in the SKILL.md frontmatter.

  • Runs code

    Shell commands in SKILL.md call:

    • gh

    From the folder's file list and the shell code blocks in SKILL.md.

  • Network

    No URLs in SKILL.md. Its commands use gh, which can reach the network depending on how they are called.

    From URLs in SKILL.md, links to its own repository left out.

  • Credentials

    Names no API keys, tokens, secrets or passwords.

    From names ending in _API_KEY, _TOKEN, _SECRET, _KEY or _PASSWORD in SKILL.md.

Context cost

Ops Desk loads about 2.9k tokens when it runs. Until then it costs about 27 tokens; SKILL.md has 932 words of instructions outside code blocks.

Always · name and description, kept in context so the agent knows when to use it
~27
When it runs · the whole SKILL.md, loaded when a task matches
~2.9k

Estimates: characters ÷ 4, the usual rule of thumb; real counts depend on the model's tokenizer. Scripts and assets cost tokens only if the agent reads them.

Safety

Auto-check: notes

The automated check noted patterns worth knowing about, such as sudo or a known installer.

  • NotePre-approves every shell command (allowed-tools: Bash)SKILL.md
    allowed-tools: Bash, Read, Grep, Glob, Skill, Agent, Workflow, AskUserQuestion, TeamCreate, SendMessage, TaskCreate

Automated static check — not a guarantee. Review scripts before installing. It scans the text of SKILL.md for risky patterns (piping downloads into a shell, reading credential files, hidden Unicode, destructive commands); files beside SKILL.md are not scanned.

SKILL.md

The full file from Lifecycle-Innovations-Limited/claude-ops at commit ee9c784, republished under its MIT licence (© Lifecycle-Innovations-Limited). 932 words, ~2,917 tokens.

Download SKILL.mdSave it as .claude/skills/ops-desk/SKILL.md (or your agent's skills folder).
name
ops-desk
description
OPS on-demand: This skill should be used when the user asks to "desk sweep", "open decisions", or…
allowed-tools
Bash, Read, Grep, Glob, Skill, Agent, Workflow, AskUserQuestion, TeamCreate, SendMessage, TaskCreate, TaskUpdate, TaskList, CronCreate, mcp__gog__gmail_search, mcp__gog__gmail_read_thread, mcp__whatsapp__list_messages, mcp__whatsapp__archive_chat, mcp__linear__list_issues, mcp__linear__get_issue, mcp__linear__update_issue
argument-hint
[scope: all|<board/company>|explicit item list]

OPS ► DESK SWEEP

Load ops-rules before acting. Public repo (no personal data). Outbound: one draft → one approval → one send. If AskUserQuestion / Workflow are missing, follow Rule 10 in ops-rules (Hermes: numbered options / two-turn Telegram card; delegate_task).

Goal: everything sitting on the owner's desk — pending decisions, staged drafts awaiting approval, unpaid invoices to chase, documents to sign, blocked issues needing an owner action — becomes a ranked queue of ready-to-approve action packages, then gets worked down one item at a time. The success metric is the same spirit as inbox-zero: an empty desk, with every remaining item either executed, decided, or parked with an explicit reason and a reminder.

When to use ops-desk vs ops-inbox

  • /ops:ops-inbox — fresh inbound comms (who wrote me, what needs a reply).
  • /ops:ops-desk — the standing backlog (what am I blocking, what decision is waiting on me, what draft is staged, what must I sign/pay/approve). Run it after an inbox pass, or whenever the owner says "what's left / get this off my desk".

Step 1 — Build the item list (unless passed explicitly in ``)

  1. Task tracker (SSOT). Whichever the box uses, in order of preference:
    • Paperclip — paperclip issue list -C <company-id> for each company in ${CLAUDE_PLUGIN_DATA_DIR}/preferences.json → paperclip.companies[], falling back to paperclip company list. Keep status=blocked|todo|in_progress issues whose title/description implies an OWNER action — matcher: decision, owner decision, approve, sign, pay, chase, staged, awaiting owner, Rule-6.
    • Linear — mcp__linear__list_issues filtered to the owner as assignee, when Paperclip is absent.
    • GitHub — gh issue list --assignee @me as last resort.
  2. Inbox residue. Genuine NEEDS_REPLY plus todo/action-labeled emails surviving the most recent /ops:ops-inbox pass (run bin/ops-inbox-scan with OIS_NO_REFRESH=1 if one already ran this session).
  3. Dedupe against items already handled this session, and collapse duplicate issues tracking the same underlying ask (mirror boards, Linear↔Paperclip twins).
  4. Shape each item as {key, refs: [issue identifiers], hint} where hint states the concrete question to answer AND what must be verified (never trust the issue title alone).

Step 2 — Run the gather workflow (READ-ONLY, batched)

Invoke the Workflow tool with the script below and args = [items]. Hard rules the script encodes:

  • Agents are READ-ONLY. No sends, archives, payments, signatures, mark-reads, or tracker mutations from inside the workflow — they read, verify, and prepare only. All execution happens later in the main session under the outbound-approval gate.
  • Batched fan-out (3 at a time). A wide parallel burst can rate-limit (429) the model-account pool — pooled CLIProxyAPI setups especially. Batches of 3 finish nearly as fast and never starve the rest of the fleet.
  • Verification over trust. Every load-bearing claim in a package must come from an actual read (issue body, thread, sqlite row) with dates/ids cited in status_summary. An item that turns out already handled comes back as action_type=already_done with evidence.
  • Partial failure → resume. If some agents die (rate limits, transient errors), fix nothing and relaunch with resumeFromRunId — completed agents replay from cache.
js
export const meta = {
  name: 'desk-sweep',
  description: 'Fan out read-only context agents over open decision/action items and return ready-to-approve action packages',
  phases: [
    { title: 'Gather', detail: 'one read-only agent per item, batches of 3' },
    { title: 'Synthesize', detail: 'order into an approval queue: quick wins first' },
  ],
}

const ACTION_SCHEMA = {
  type: 'object', additionalProperties: false,
  required: ['key', 'title', 'status_summary', 'recommendation', 'action_type'],
  properties: {
    key: { type: 'string' },
    title: { type: 'string' },
    status_summary: { type: 'string', description: '2-4 sentences: what this is, current state, evidence (dates/ids)' },
    recommendation: { type: 'string', description: 'single concrete next action with an approvable default' },
    action_type: { type: 'string', enum: ['send_email', 'send_whatsapp', 'decision', 'sign', 'pay', 'chase', 'info_only', 'archive', 'already_done'] },
    draft: { type: ['object', 'null'], additionalProperties: false, properties: {
      channel: { type: 'string', enum: ['email', 'whatsapp'] },
      to: { type: 'string' }, cc: { type: 'string' }, subject: { type: 'string' },
      body: { type: 'string', description: "FULL final text in the owner's voice + thread language" },
      reply_to_message_id: { type: 'string' },
    }},
    decision_options: { type: 'array', items: { type: 'string' } },
    blockers: { type: 'array', items: { type: 'string' } },
    effort_minutes: { type: 'number' },
    confidence: { type: 'string', enum: ['high', 'medium', 'low'] },
  },
}

phase('Gather')
const ITEMS = (typeof args === 'string' ? JSON.parse(args) : args) || []
if (!ITEMS.length) throw new Error('desk-sweep needs args: [{key, refs, hint}]')

const CHUNK = 3
const packages = []
for (let i = 0; i < ITEMS.length; i += CHUNK) {
  const batch = ITEMS.slice(i, i + CHUNK)
  const res = (await parallel(batch.map(it => () =>
    agent(
      `You are a READ-ONLY desk-sweep context agent. You MUST NOT send, archive, pay, sign, ` +
      `mark-read, or mutate ANYTHING — no mail sends, no POSTs to send/archive endpoints, no ` +
      `tracker mutations. Read, search, and prepare only.\n\n` +
      `ITEM key="${it.key}": ${it.hint}\nReferences: ${JSON.stringify(it.refs || [])}\n\n` +
      `Gather context read-only, every shell command with an explicit timeout ` +
      `("timeout --signal=TERM --kill-after=5s 30s <cmd>"): the task tracker record(s) for each ` +
      `ref incl. comments; Gmail via gog (search then thread get -j; messages nest under ` +
      `.thread.messages); the WhatsApp bridge store read-only via sqlite3 (merge a person's @lid ` +
      `and @s.whatsapp.net chats; resolve names via contacts); local files if referenced.\n\n` +
      `RULES: verify every load-bearing claim with an actual read; if already handled, return ` +
      `action_type=already_done with evidence. Drafts must be the FULL final text in the owner's ` +
      `voice and the thread's language, no AI tone. Decisions get 2-4 concrete options, ` +
      `recommended first. Your final output MUST be the structured action package.`,
      { label: `gather:${it.key}`, phase: 'Gather', schema: ACTION_SCHEMA }
    )
  ))).filter(Boolean)
  packages.push(...res)
  log(`gathered ${packages.length}/${ITEMS.length}`)
}

phase('Synthesize')
return await agent(
  `You are READ-ONLY. Order these action packages into an approval queue: quick wins first ` +
  `(low effort, high confidence, drafts ready), then decisions, then heavy/blocked. Drop ` +
  `nothing; flag conflicts between packages.\n\n` + JSON.stringify(packages, null, 2) +
  `\n\nReturn {queue: [same packages verbatim + "rank"], conflicts: [], stats: {total, ` +
  `ready_drafts, decisions, already_done}}.`,
  { label: 'synthesize', phase: 'Synthesize',
    schema: { type: 'object', additionalProperties: true, required: ['queue'] } }
)
Show full SKILL.md (459 more words)Show less

Step 3 — Work the approval queue (main session, outbound gate)

Per package, in rank order:

  • already_done / info_only → report in one line and close the item in the task tracker (status → done). This is tracker-only closure: no mailbox, chat or file is archived, moved or marked read. These are the free wins.
  • archive → a package from a read-only agent is a recommendation, not owner authorization. Show the item and the evidence, then archive it only after explicit owner authorization for that item (e.g. via AskUserQuestion). Without it, leave the thread where it is and report the item as still open.
  • send_email / send_whatsapp / chase → exactly as in /ops:ops-inbox and ops-rules Rule 6: one draft, one approval, one send, only through the host's approved outbound gate. The gate shows the exact recipient and full latest text; approval is the owner's native yes bound to those bytes, never a token, counter or earlier approval. Re-read the live thread before sending, verify delivery on a fresh read, then update the tracker. Archive only under explicit archive authorization. If the gate refuses or does not support the channel, keep the item owned with the exact diagnostic; never use a direct CLI, bridge or API sender.
  • decision → present the options (recommended first) via AskUserQuestion, record the outcome in the tracker, execute any follow-through it implies.
  • sign / pay → prepare only: surface the exact link/document/amount and what to check before signing/paying. NEVER sign or move money autonomously (Rule 5).

Step 4 — Wrap up

Report: sent / decided / closed / still-open-with-reason. Schedule reminders (CronCreate, non-outbound) for anything parked with a follow-up horizon. The task tracker is the SSOT — every touched item gets its status updated there before the summary prints.

Agent Teams support

If CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 is set, use Agent Teams when fanning out the gather workflow across independent item sources (Paperclip per company, Linear, GitHub, Gmail/WhatsApp residue, etc.) in parallel. This enables:

  • Gather agents share context mid-flight (one surfaces a high-priority payment → the ranker can prioritize before the desk-sweep plan is emitted)
  • You can steer scope in real time ("skip Linear, focus on Paperclip only")
  • Agents report progress as each gather round completes, so the main session can start the approval-queue step earlier on the items already returned

Team setup (only when the flag is enabled, and only for genuinely parallel sources — a single-source desk sweep stays inline):

TeamCreate("ops-desk")
Agent(team_name="ops-desk", name="paperclip-gather", prompt="Gather owner-action items across all configured Paperclip companies")
Agent(team_name="ops-desk", name="linear-gather", prompt="Gather Linear items assigned to the owner")
Agent(team_name="ops-desk", name="inbox-residue", prompt="Gmail/WhatsApp NEEDS_REPLY + todo-labeled items from the last inbox pass")
Agent(team_name="ops-desk", name="github-chases", prompt="Open PRs awaiting owner review, blocked issues")

After the team reports, the main session runs Step 3 (work the approval queue) on the merged, ranked list. Each SendMessage from a gather agent is a structured action package — never an outbound send (Rule 6: per-draft outbound approval gate stays in the main session, not in the team).

If the flag is NOT set, fall back to standard fire-and-forget Agent subagents (one per source) — same shape, slower wall-clock because no shared context.

© Lifecycle-Innovations-Limited, MIT. Rendered from Markdown: HTML in the file is shown as text, images as links, and headings moved down two levels. Raw file

Files

Just SKILL.md in claude-ops/skills/ops-desk of Lifecycle-Innovations-Limited/claude-ops.

Open the folder on GitHubat commit ee9c784

Compare with similar skills

Ops Desk next to the 5 skills that share the most tags, products or categories with it. Stars are the repository's; “used in” counts other GitHub owners with a copy.

Ops Desk compared with similar skills
SkillStarsUsed inTokensAuto-checkLicenceRepo updated
Ops Desk this skillLifecycle-Innovations-Limited/claude-ops542—~2.9kAutomated safety check: NotesMIT
Garden Inboxpaperclipai/paperclip99k—~1.1kAutomated safety check: PassMIT
Continuetelegramdesktop/tdesktop33k2 repos~9.4kAutomated safety check: PassGPL-3.0
Process Inboxtelegramdesktop/tdesktop33k1 repos~5.4kAutomated safety check: PassGPL-3.0
Process InboxTDesktop-x64/tdesktop3k—~4.3kAutomated safety check: PassGPL-3.0
Career-Ops Gmail Lead Plugincareer-ops-hq/career-ops74k—~233Automated safety check: NotesMIT

Similar skills

  • Garden Inbox

    paperclipai/paperclip

    Scan a Paperclip user's Mine inbox, classify reversible archive candidates, request checkbox confirmation, and archive only accepted selections.

    99k GitHub stars~1.1k tokensUpdated today
    Productivity & AutomationAuto-check passed
  • Continue

    telegramdesktop/tdesktop

    Continue autonomous Telegram Desktop development from the shared ai-tdesktop repository.

    33k GitHub starsUsed in 2 repos~9.4k tokens
    Productivity & AutomationAuto-check passed
  • Process Inbox

    telegramdesktop/tdesktop

    Process the local ignored ai-tdesktop inbox into durable, independently testable Telegram Desktop task records while task execution worktrees remain active.

    33k GitHub starsUsed in 1 repo~5.4k tokens
    Productivity & AutomationAuto-check passed
  • Process Inbox

    TDesktop-x64/tdesktop

    Process the local ignored ai-tdesktop inbox into durable, independently testable Telegram Desktop task records while task execution worktrees remain active.

    3k GitHub stars~4.3k tokensUpdated 18 days ago
    Productivity & AutomationAuto-check passed
  • Career-Ops Gmail Lead Plugin

    career-ops-hq/career-ops

    Pulls job leads from a Gmail label into the career-ops pipeline, extracting job URLs from DMARC-passing emails and de-duplicating against existing leads.

    74k GitHub stars~233 tokensUpdated today
    Productivity & AutomationAuto-check: notes
  • Atomicmail

    Atomic-Mail/atomic-mail-agentic

    Read and write email through the Atomic Mail from an AI agent.

    266 GitHub starsUsed in 1 repo~2k tokens
    Productivity & AutomationAuto-check passed

More from Lifecycle-Innovations-Limited/claude-ops

All 67 skills in this repo
  • Ops Dash

    Lifecycle-Innovations-Limited/claude-ops

    OPS on-demand: This skill should be used when the user asks to "ops dashboard", "pixel HQ", or…

    542 GitHub stars~4.7k tokensUpdated today
    Auto-check: notes
  • Ops Gtm

    Lifecycle-Innovations-Limited/claude-ops

    OPS on-demand: This skill should be used when the user asks to "go to market", "GTM plan", or…

    542 GitHub stars~3.7k tokensUpdated today
    Auto-check: notes
  • Ops Marketing

    Lifecycle-Innovations-Limited/claude-ops

    OPS on-demand: This skill should be used when the user asks to "klaviyo", "ads spend", or…

    542 GitHub stars~3.5k tokensUpdated today
    Auto-check: notes
  • Ops Socials

    Lifecycle-Innovations-Limited/claude-ops

    OPS on-demand: This skill should be used when the user asks to "tweet", "post to linkedin", or…

    542 GitHub stars~3.6k tokensUpdated today
    Auto-check: notes
  • Ops Yolo

    Lifecycle-Innovations-Limited/claude-ops

    OPS on-demand: This skill should be used when the user asks to "yolo mode", "run the business today"…

    542 GitHub stars~4k tokensUpdated today
    Auto-check: notes
  • Ops Monitor

    Lifecycle-Innovations-Limited/claude-ops

    OPS on-demand: This skill should be used when the user asks to "datadog", "APM alerts", or…

    542 GitHub starsUsed in 1 repo~1.5k tokens
    Auto-check: notes

Questions about Ops Desk

What does Ops Desk do?

OPS on-demand: This skill should be used when the user asks to "desk sweep", "open decisions", or…. Ops Desk is an agent skill from Lifecycle-Innovations-Limited/claude-ops.

When should I use Ops Desk?

Ops Desk fits situations like: asks to desk sweep; tasks that involve Email management.

How do I install Ops Desk in Claude Code?

Run `npx skills add Lifecycle-Innovations-Limited/claude-ops --skill ops-desk -a claude-code`. Or copy the skill folder (claude-ops/skills/ops-desk in Lifecycle-Innovations-Limited/claude-ops) into .claude/skills/ops-desk in your project. Claude Code loads it when a task matches its description.

How do I install Ops Desk in Codex?

Run `npx skills add Lifecycle-Innovations-Limited/claude-ops --skill ops-desk -a codex`. Or copy the skill folder (claude-ops/skills/ops-desk in Lifecycle-Innovations-Limited/claude-ops) into .agents/skills/ops-desk in your project. Codex loads it when a task matches its description.

Can I use Ops Desk in Cursor, Gemini CLI or GitHub Copilot?

Cursor, Gemini CLI, GitHub Copilot and OpenCode also load SKILL.md folders. With the skills CLI, run `npx skills add Lifecycle-Innovations-Limited/claude-ops --skill ops-desk -a cursor` (or -a gemini-cli, github-copilot or opencode for the others). To copy it by hand, put the folder in .cursor/skills/ops-desk, .gemini/skills/ops-desk, .github/skills/ops-desk and .opencode/skills/ops-desk in your project.

What does Ops Desk need to run?

Going by SKILL.md and its folder, Ops Desk needs the command-line tools its instructions call (gh). Its frontmatter pre-approves these tools: Bash, Read, Grep, Glob, Skill, Agent, Workflow, AskUserQuestion, TeamCreate, SendMessage, TaskCreate, TaskUpdate, TaskList, CronCreate, mcp__gog__gmail_search, mcp__gog__gmail_read_thread, mcp__whatsapp__list_messages, mcp__whatsapp__archive_chat, mcp__linear__list_issues, mcp__linear__get_issue, mcp__linear__update_issue.

Does Ops Desk access the network?

SKILL.md contains no URLs. Its commands use gh, which can reach the network depending on how they are called. This is read from the text; nothing was executed.

Is Ops Desk safe to install?

Our automated static check of SKILL.md found notes only (pre-approves every shell command (allowed-tools: bash)), nothing it rates as a warning. It is not a guarantee. Review the folder before installing.

What licence does Ops Desk use?

Ops Desk is published under the MIT licence (the repository's licence). It allows redistribution, so the full SKILL.md is shown on this page.

How many tokens does Ops Desk use?

About 2.9k tokens (SKILL.md is roughly 12k characters). Agents keep only the skill's name and description in context until a task matches; then they load SKILL.md in full.

What are the alternatives to Ops Desk?

Skills that share tags, products or a category with Ops Desk: Garden Inbox (paperclipai/paperclip, 99k stars), Continue (telegramdesktop/tdesktop, 33k stars), Process Inbox (telegramdesktop/tdesktop, 33k stars) and Process Inbox (TDesktop-x64/tdesktop, 3k stars). The comparison table on this page puts their stars, adoption, token cost, safety result and licence side by side.

Who maintains Ops Desk?

Lifecycle-Innovations-Limited (a GitHub organization) maintains it in Lifecycle-Innovations-Limited/claude-ops, which has 542 GitHub stars. The repository holds 67 skills in this directory. The repository was last updated on October 10, 2026.

Source: Lifecycle-Innovations-Limited/claude-ops on GitHub. Facts on this page come from the repository at the commit we read; the author's words are quoted as theirs.