proposal(198): Self-Host & Dependency Currency#2883
Conversation
Surfaced 2026-06-18/19 from the PR #2882 review: the Specrew self-host had drifted ~6 releases and fallen behind Spec Kit (0.10.0 removed --ai, which specrew init still passes). Proposes deterministic regeneration + release-time self-update + per-release dependency currency + a CI drift gate. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
Pull request overview
Adds Proposal 198 (“Self-Host & Dependency Currency”) to document a release/CI governance approach for keeping Specrew’s self-hosted generated assets and platform dependency pins current, and updates the proposal index to reference it.
Changes:
- Add
proposals/198-self-host-currency.mdwith the candidate proposal content and frontmatter. - Update
proposals/INDEX.mdto reflect the new candidate count and include Proposal 198 in the index.
Reviewed changes
Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.
| File | Description |
|---|---|
| proposals/INDEX.md | Updates the candidate section count and adds the Proposal 198 row to the index. |
| proposals/198-self-host-currency.md | Introduces the full Proposal 198 candidate write-up (motivation, parts, risks, and composition map). |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 3a04566701
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| - **Broken beta breaks the self-host.** Beta-track regeneration can wedge the dev environment. Mitigate via revertable regeneration + validation-then-promote ordering. | ||
| - **190 interaction.** An automated governance-surface rewrite is exactly what 190 guards against; shipping without the reconciliation either trips the guard or weakens it. Mitigate by treating the 190 reconciliation as a hard dependency, not a footnote. | ||
| - **Per-release diff churn.** Mitigate via deterministic generation (Part 1) so the only diffs are real version deltas. | ||
| - **Adoption friction.** A hard drift gate that fires before generation is fully deterministic would block unrelated PRs. Mitigate via warn-then-block phasing gated on Part 1 landing. |
There was a problem hiding this comment.
Add the missing required proposal sections
For a new proposal, docs/methodology/proposal-discipline.md says every proposal must include populated How, Acceptance criteria, Out of scope, Composition, Risks, Status history, and Cross-references, and the reviewer checklist specifically calls this out for new proposals. This file ends after Risks without acceptance criteria, out-of-scope boundaries, status history, or cross-references, so it is not actionable/traceable under the documented proposal discipline and should be rejected by review until those sections are added.
Useful? React with 👍 / 👎.
Candidate proposal surfaced from the #2882 review.
Why: the Specrew repo's own deployed assets had drifted ~6 releases (extension registration stuck at 0.32.0,
AGENTS.md/CLAUDE.mdabsent, the Feature 177 skill never deployed), and Specrew had fallen behind Spec Kit (max_tested0.9.0 vs upstream 0.11.2 -- 0.10.0 removed--ai, whichspecrew initstill passes) -- because no release or CI step regenerates the self-host or bumps dependencies.What: four parts -- deterministic + lint-clean generation (prerequisite), a release-time self-update folded into beta-validation, a per-release dependency-currency step (additive bumps fold into the release, breaking bumps spin out as
platform-upgradeslices), and a CI drift gate. Reconciles with Proposal 190.Full text:
proposals/198-self-host-currency.md. Composes 060/061/075/132/173/190.🤖 Generated with Claude Code