Skip to content

Write the GitHub feedback re-entry RFC #95

Description

@markfreeman-dot-dev

v2 migration #100M7-lifecycle (epic #15) · Size S · Architecture

Goal

An RFC treating GitHub review mechanics as structured re-entry: PR-per-cell, changes-requested comments become a focused_directive, merge unblocks dependents.

Context

Petri's human touchpoints are CLI prompts today; the focused-directive mechanism already exists on the weakest_link path. The mapping raises real questions worth a document first: cell-to-PR identity, auth, poll vs webhook, and merge vs mechanical convergence.

Scope

In:

  • RFC in docs/, linked from the ARCHITECTURE-V2 ADR index: motivation, non-goals, and cell-to-PR mapping options (branch-per-cell vs PR-per-colony vs labels).
  • Event model: github_feedback_received domain events carrying the focused-directive payload, registered in EVENT_DATA_MODELS.
  • Directive-extraction rules (which comment content becomes the directive, how multi-reviewer disagreement resolves) and merge-unblock semantics (flag-don't-requeue default; merge is explicit human requeue).
  • Auth model (gh CLI vs token vs GitHub App) and the poll-vs-webhook tradeoff for a tool with no always-on server.
  • Invariant stated outright: GitHub approval never substitutes for mechanical convergence; plus a sequence diagram and proposed follow-on issue split.

Out:

  • No code, GitHub App registration, or dashboard integration.
  • No implementation scheduling (a future milestone decides).

Acceptance criteria

  • RFC merged in docs/ and linked from the ARCHITECTURE-V2 ADR index.
  • Covers cell-to-PR mapping, directive extraction, merge-unblock semantics, auth, and poll-vs-webhook, each with a recommended option and rationale.
  • States the invariant that GitHub approval never bypasses mechanical convergence.
  • Proposes the follow-on implementation issue split with rough sizes.
  • Explicitly labeled design-first: no implementation is scheduled by this RFC.

Pointers


#N in this issue refers to entries in docs/field-reports.md, not issues in this repo.
To claim this issue, leave a comment. Questions welcome — see CONTRIBUTING.md.

Metadata

Metadata

Assignees

No one assigned

    Labels

    documentationImprovements or additions to documentationlifecycleRe-decomposition, feed, scan, analystmigration-v2Part of the v2 migration backlogrfcDesign-first; discussion expected before implementationsize:SRough effort: hours

    Type

    No type

    Projects

    Status
    Todo

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions