Skip to content

docs: list context workflows - #514

Merged
pdettori merged 1 commit into
rossoctl:mainfrom
moonlight16:docs/context-workflows
Oct 14, 2026
Merged

pdettori merged 1 commit into
rossoctl:mainfrom
moonlight16:docs/context-workflows

Conversation

@moonlight16

Copy link
Copy Markdown
Member

Part of the work split out of #455 (the original context-delegation PR), which kept a workflows table inside the delegate skill.

Adds docs/context-workflows.md: a living list of the ways we expect people to move context into and out of Moca. Ten workflows in three groups (delegate work from a laptop, keep context in Moca, move context between places), each with a short flow sketch, the sandbox and storage it uses, and its status, linked to #482, #494, #476 (phase 2: per-workload sandboxes and storage), #358 (per-user PVC authorization) and #495 (one CLI for delegating subagent tasks).

The descriptions will keep improving; this gets the list into the repo so we can point at it and refine it in follow-up PRs.

🤖 Generated with Claude Code

A living list of the ways we expect people to move context into and out of Moca: what sandbox and storage each workflow uses, and where it stands. Centralizes the workflows table that lived in the delegate skill in rossoctl#455.

Assisted-By: Claude (Anthropic AI) <noreply@anthropic.com>
Signed-off-by: Jeremy Cohn <Jeremy.Cohn@ibm.com>

@ingpaolodettori-dev ingpaolodettori-dev left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Useful list, and the statuses check out against main: #482 and #494 are merged, #476, #358 and #495 are open, and contextctl ctx backup exists in rossoctl/context-service. No blockers. The inline comments are about where the page lives, a few places where it could be more precise about what phase 1 actually gives you, and some wording.

Author: moonlight16 (MEMBER — maintainer)
Areas reviewed: Docs (checked against docs/CONTRIBUTING-DOCS.md, docs/context-service.md, the #494 description, and rossoctl/context-service README/docs)
Agent/IDE config (.claude/.vscode): none
Commits: 1 commit, all signed-off: yes
CI status: passing (incl. docs / check-links)

Comment thread docs/context-workflows.md
data that the task works from. This page lists the ways we expect people to move context into and
out of Moca, and what each one needs. It is a living list; propose changes in a PR.

A **workload** is an environment for delegated tasks: where they run and the context they work from

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

suggestion: docs/CONTRIBUTING-DOCS.md ("Where things go") has no home at the root of docs/, and nothing links to this page. docs/index.md doesn't list it either, so it starts out as one of the orphans that REORG-PLAN is cleaning up. Consider moving it to docs/concepts/, or to docs/notes/ if it's meant as a living working list, and adding a row to docs/index.md. A maturity line would also help (e.g. that it's a planning list rather than a description of shipped behavior), since CONTRIBUTING-DOCS asks every page for one.

Comment thread docs/context-workflows.md

A **workload** is an environment for delegated tasks: where they run and the context they work from
([docs/context-service.md](context-service.md)). Today a workload belongs to the user who created
it. Context is stored by [Context Service](https://github.com/rossoctl/context-service).

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: CONTRIBUTING-DOCS.md (Terminology) says the product name is MOCA. This page uses "Moca" throughout. context-service.md does too, so this is optional, but new pages should use the documented spelling.

Comment thread docs/context-workflows.md

### Keep context in Moca

| # | Workflow | Flow | When to use it | Sandbox | Storage | Status |

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

suggestion: "Server done" is true, but phase 1 has limits a reader planning on workflows 1 and 2 will want to know about (from #494 and docs/context-service.md): workload runs are synchronous only (async /runs with a workloadId returns 501), they need a logged-in user, and /workloads returns 501 workloads_unavailable unless MOCA_CONTEXT_WORKLOADS_ENABLED=1 and CONTEXT_SERVICE_URL are set. A short footnote in the Status section, or a link to the context-service.md limits, would cover it.

Comment thread docs/context-workflows.md
## Sandbox and storage types

Sandboxes:

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: The rossoctl/context-service README already lists "Automatic capture and restore of native agent-harness state" for Claude Code, Codex, OpenCode and Pi, and it has docs/context-portability.md. If the missing part is cross-harness translation (Claude Code → Codex), say that and link the existing capture/restore docs. As written, "No design yet" undersells what's there.

Comment thread docs/context-workflows.md

Sandboxes:

- **Shared:** the operator's shared, statically provisioned sandboxes. Tasks from many workloads

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: The workflow cross-references (feeds #5 here, Unlike #2 on line 95, as in #5 on line 102) look the same as the issue references (#476, #358) on the same page. Using "workflow 5" / "workflow 2" would avoid the confusion, and also any autolinking when this text is quoted in issues or PRs.

Comment thread docs/context-workflows.md
- **Either:** the workflow works with both.

Storage:

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

suggestion: Please add that shared sandboxes are not a security boundary between users. #494 says so explicitly: tasks from different users can run on the same sandbox, as the same Linux user, one after another, and read-only is enforced only by file permissions. Row 3 implies this already ("without seeing each other" needs per-workload sandboxes), but stating it here keeps readers from assuming workflows 1 and 2 isolate users.

Comment thread docs/context-workflows.md
tasks build up, such as cloned repositories.
- **Existing Context PVC:** a Context that already exists, used instead of a new upload, and
authorized by Context Service's per-user permissions rather than a PVC name.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: "Delegated tasks always mount it rather than copying from it" reads as if it contradicts rows 1 and 2, where delegated tasks get a copy. Maybe: "In per-workload sandboxes (#476), delegated tasks mount it rather than copying from it."

Comment thread docs/context-workflows.md
**5. Reuse a remote dataset.** Upload the context once and keep it. Later workloads, in other
batches or days later, point at the same Context instead of uploading again. Unlike #2, the
Context outlives the workload that first used it.

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: "The BugStone pipeline project" isn't defined anywhere a reader of this page would find it (only in passing in two specs). Either gloss it in a few words or drop the sentence.

@pdettori
pdettori merged commit 8c1b6b1 into rossoctl:main Oct 14, 2026
14 checks passed
moonlight16 added a commit that referenced this pull request Oct 14, 2026
Move the context workflows list to docs/concepts/ with a Preview line and index entries, use the MOCA spelling, state phase 1 limits and that shared sandboxes are not a security boundary, link Context Service's capture/restore for workflow 9, and say "workflow N" for cross-references. Also correct that under SH_REQUIRE_AUTH an unauthenticated workload run is 401 token_required first (missed #516).

Assisted-By: Claude (Anthropic AI) <noreply@anthropic.com>
Signed-off-by: Jeremy Cohn <Jeremy.Cohn@ibm.com>
moonlight16 added a commit that referenced this pull request Oct 14, 2026
docs: address #514 review; workload runs need a token first
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants