Skip to content

Latest commit

 

History

History
70 lines (53 loc) · 4.3 KB

File metadata and controls

70 lines (53 loc) · 4.3 KB

Workflows: choosing how a task runs

Every task runs a workflow: the lifecycle that decides what states the task moves through, who advances each one (you or the agent), what the agent must finish before it can move on (its responsibilities), and which extra skills the agent has in its container. Picking a workflow is how you set the line between what the agent may do on its own and what needs your sign-off.

You choose a workflow when you create a task. On the dashboard press n, pick the repo, then pick the workflow. This page is the catalog; each workflow has its own page with the details. For the task itself — the object a workflow drives, its properties, and its lifecycle — see Tasks.

The built-in workflows

Workflow What it does, and when to use it Where it ships
github-peer-reviewed Ships a GitHub PR that a peer reviews before it merges. Use for GitHub changes that need a second person's sign-off. A GitHub PR (peer-approved)
github-self-reviewed Ships a GitHub PR that you review yourself, with no peer-review gate. A GitHub PR (self-approved)
local-git-self-reviewed Keeps the work local: commits to a branch and merges it, with no GitHub, PR, or CI. Use when the change never leaves the machine. A local branch merged into the base
spike Open-ended agent work with no gates. Use for exploration, debugging, and research, until you call it done. Nothing lands on its own
orchestrator An agent that decomposes a goal into child tasks, each pre-planned and handed to you ready to approve. Use to fan work out across agents. New pre-planned child tasks
setup-repo A host-side setup utility (no container) that mints a repo's claude auth token. Launched from the repos screen, not the task picker. A token in the repo's env-file

Any task can also be dropped at any time (dashboard x), which moves it to DROPPED without merging or shipping anything.

The planning step

Four change-making workflows (github-peer-reviewed, github-self-reviewed, local-git-self-reviewed, and, per child, orchestrator) start in PLANNING. Before the agent can leave that state it must:

  • Write a plan as the task's plan.md artifact. This is your chance to redirect before any code is written.
  • Record a token estimate so the task's projected cost is tracked.

The plan is a task artifact, which you read from the dashboard: highlight the task and press a to open its plan.md. (Artifacts are the task's own documents, kept with the task rather than in the repo, so the dashboard is where you read them.)

You approve the plan by advancing the task out of PLANNING (attach with t, run /advance). spike and setup-repo have no planning step.

Who advances a state

Each state is advanced by either you or the agent:

  • You advance the foreground states (plan approval, sign-off before merge). The agent fills in its responsibilities and then waits; nothing proceeds until you say so. Attach with t and run /advance to approve, and the agent starts a fresh turn.
  • The agent advances the background states (merging). Once its responsibilities are met (the PR is merged, or the branch is merged locally) it moves the task on by itself.

How workflows are offered

  • Default-on vs. opt-in. spike and orchestrator are shown for every repo by default. github-peer-reviewed, github-self-reviewed, and local-git-self-reviewed are opt-in: enable them per repo (in the repo's workflow settings) before they show up in the task-creation picker.
  • Hidden utilities. setup-repo is hidden from the pickers entirely; you launch it from the repos screen's setup hotkey, not by creating an ordinary task.

Adding your own

Workflows are just Workflow subclasses. Drop a module defining one into ~/.config/panopticon/workflows/ (or point the session service at a directory with --workflows-path) and it registers automatically, with no change to the core service. See ADR 0004 and docs/design/ARCHITECTURE.md §7 for the authoring model.