Skip to content

feat(plan-work-items): classify each item's implementation, review, and expected paths#111

Draft
taminomara wants to merge 1 commit into
testdouble:mainfrom
taminomara:feat/plan-work-items-item-classification
Draft

feat(plan-work-items): classify each item's implementation, review, and expected paths#111
taminomara wants to merge 1 commit into
testdouble:mainfrom
taminomara:feat/plan-work-items-item-classification

Conversation

@taminomara

Copy link
Copy Markdown
Contributor

This is a draft MR with diff to plan-work-items cherry-picked from my #96 WIP branch.

The idea for the change: plan-work-items gains knowledge about which skill or agent can be used to implement and review a particular work item. It also knows whether that skill/agent can finish work unattended, or if it's explicitly interactive. Based on this, each item gains classification: build AFK/HITL, and review AFK/HITL. Additionally, it gains a separate classification on whether an item requires making HITL decisions before it can be implemented (i.e. deferred OIs or assumptions that require human intervention before build starts).

On top of it, it also adds work item types: deliverable is a regular "write code, then review"; audit is "verify something works and produce a report" (can be good for deferring OIs/assumptions into implementation, or adding verification that can't be done via tests); spike is for research or investigation items. Note that audit and spike are separate due to initial implementation constraints that I've eventually solved, so I'll merge them into one type before moving PR into ready state.

The change is not yet ready for deep review or merge, there are things to polish, edge cases to debug, and also I need to look at it with fresh eyes after working on something else for a bit. But you're welcome to comment on the general idea and approach. Or, well, you can review it, just know that this is not the final version.

…nd expected paths

Enrich the producer so each work item carries the markers the downstream
driver needs, without pulling in the driver itself:

- Add a Type field (deliverable / audit / spike) and rename Tests -> Checks
  in the work-item template.
- Add Suggested implementation, Suggested review, Requires pre-work decisions,
  and Expected paths fields, selected via a new deliverable-skill catalog.
- Direct project-manager to derive those markers and expected paths, and close
  the run with the three-bucket (unattended / pauses-for-human) recommendation.
- Reframe the upstream unverified-assumptions guidance onto Requires pre-work
  decisions, since item-level HITL/AFK moves to the per-marker model.

Split out of feat/autonomous-driver-core-loop as a standalone producer-side
change; carries no implement-work-items edits.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01BvgX88kC9h9xbe4zGLf7BC
@taminomara

taminomara commented Jul 10, 2026

Copy link
Copy Markdown
Contributor Author

Oh, also none, HITL wording is super confusing, I'll change it. Sometimes it means "work with user interactively on this, no specific skill/guidance fits", and sometimes it means "review is not needed, skip", and other times its "user reviews manually".

Also also, Expected paths is something I've decided to drop in favor of a dedicated scope-check stage in implementation driver.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant