agentleFS
OpenCovenOpenCoven/coven/skills/lobster/SKILL.md

lobster

Skill75 starsChanged 4 months agoMITView on GitHub

Run deterministic typed pipelines, workflow files, and approval-gated automations via the Lobster CLI. Use for multi-step automations that need safety (approval gates), statefulness (cursors/checkpoints), and token efficiency (one call vs many LLM re-plans). Triggers on workflow automation, pipeline, deterministic execution, approval gate, email triage, PR monitor.

1 other project uses a version of this.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

1 rule. About 0 are common, 0 are distinctive, and 1 is very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈0

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 1

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    When status: "needs_approval", call lobster resume --token <token> --approve yes to continue.

    line 98

≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Lobster — Workflow Shell

Typed, local-first pipeline engine. Deterministic execution, approval gates, persistent state. Saves tokens by replacing multi-step LLM re-planning with one-shot pipeline calls.

CLI

lobster '<pipeline>'                    # inline pipeline
lobster run path/to/workflow.lobster    # YAML workflow file
lobster run --file wf.lobster --args-json '{"key":"val"}'
lobster resume --token <token> --approve yes|no
lobster doctor                          # health check
lobster help <command>                  # per-command help

Pipeline Commands

Data flow

CommandPurposeExample
execRun OS commandexec --json --shell 'echo [1,2]'
whereFilterwhere 'status=OPEN'
pickProject fieldspick 'title,author,url'
headFirst N itemshead --n 5
sortStable sortsort --key updatedAt --desc
groupByGroup by keygroupBy --key category
dedupeRemove dupesdedupe --key id
mapTransformmap --wrap items / map status=done
templateRender texttemplate --text '{{title}} by {{author}}'

Output

CommandPurpose
jsonJSON output
tableTable output

Control flow

CommandPurpose
approveHalt for approval (--emit for non-TTY/tool mode)
state.getRead persistent state
state.setWrite persistent state
diff.lastCompare to last snapshot

Integrations

CommandPurpose
clawd.invokeCall OpenClaw tool endpoint
gog.gmail.searchFetch Gmail threads
gog.gmail.sendSend Gmail
email.triageCategorize + draft replies
llm_task.invokeCall LLM with typed payloads
workflows.runRun named workflow
workflows.listList available workflows

Pipe syntax

Commands are piped with |. Data flows as typed JSON arrays (not text).

lobster "exec --json --shell 'gh pr list --json number,title,state' | where 'state=OPEN' | pick 'number,title' | head --n 5 | table"

Workflow files (.lobster)

YAML files with steps, env vars, conditions, and approval gates:

name: deploy-check
steps:
  - id: tests
    command: exec --json --shell 'pnpm test --json'
  - id: approve
    command: approve --prompt "Tests passed. Deploy to staging?"
    approval: required
  - id: deploy
    command: exec --shell 'pnpm deploy:staging'
    condition: $approve.approved

Run: lobster run deploy-check.lobster

Steps can reference prior outputs: stdin: $tests.stdout

Tool mode (for OpenClaw integration)

lobster run --mode tool '<pipeline>'

Returns a JSON envelope: { ok, status, output, requiresApproval }

When status: "needs_approval", call lobster resume --token <token> --approve yes to continue.

Built-in recipes

GitHub PR Monitor

lobster "workflows.run --name github.pr.monitor --args-json '{\"repo\":\"owner/repo\",\"pr\":123}'"

Tracks PR state changes with diff detection.

State persistence

state.set / state.get persist JSON values across runs (e.g., last-processed cursor). diff.last compares current data to a stored snapshot and emits only changes.

Key principles

  • One call — replace 10+ LLM tool calls with one pipeline
  • Approval gates — hard stops, not prompt hints. Pipeline cannot continue without explicit resume.
  • Stateful — cursors/checkpoints survive across runs
  • No auth surface — Lobster never owns tokens; uses env vars or OpenClaw's existing auth

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.