Skip to content

docs: record the v0.5.2 catalog landing and take the real bundle install receipt - #51

Merged
rdryfoos merged 1 commit into
mainfrom
catalog-landed
Sep 24, 2026
Merged

rdryfoos merged 1 commit into
mainfrom
catalog-landed

Conversation

@rdryfoos

Copy link
Copy Markdown
Owner

One commit, four files, all under docs/submission/. Do not merge; Rik's hand.

One deviation from the order, stated rather than done quietly. The order says base 500075e. That commit is main's state before #50 merged; main is now 7e7b4c4, and branching from 500075e would have silently dropped the ONBOARD companion-aside change. This branch is cut from 7e7b4c4 and targets main. 500075e is an ancestor of it, so nothing the order intended is lost.

1. The receipt, taken after the merges because it could not be taken before

specify bundle install specassay resolves through the community catalog, which read 0.4.12 until #4737 merged. Clean specify init project outside this repository, Spec Kit 1.0.5 installed from its v1.0.5 tag, Python 3.11.15, Linux, 2026-09-24.

The community catalog resolves both components at 0.5.2 and will not install them. policy=discovery-only is Spec Kit's own setting:

$ specify bundle info specassay
specassay v0.5.2 — SpecAssay
  Source: community (discovery-only)
  Requires Spec Kit: >=0.14.0,<2.0.0
  Components (added on install):
    extensions:
      - specassay-check v0.5.2
    presets:
      - specassay v0.5.2 (priority=10, strategy=append)
  This source is discovery-only; the bundle cannot be installed from here.

$ specify bundle install specassay
Error: Bundle 'specassay' resolves only from a discovery-only source
('community'); it cannot be installed from there.

That refusal is the reason the README's install section adds our three catalogs --install-allowed first. Run verbatim, it then installs:

$ specify bundle install specassay
Updated execute permissions on 4 script(s) recursively
✓ Installed 'specassay' (2 added, 0 already present).

$ specify bundle list
  specassay v0.5.2 (2 components, installed 2026-09-24T17:21:31Z)

$ specify preset list
  SpecAssay (specassay) v0.5.2 — enabled — priority 10 / Templates: 3

$ specify extension list
  ✓ SpecAssay Check (v0.5.2)
     Commands: 5 | Hooks: 1 | Priority: 10 | Status: Enabled

Resolved versions, confirmed on disk and not only in the listings:

resolved on disk
bundle specassay 0.5.2 .specify/bundle-records.json
extension specassay-check 0.5.2 .specify/extensions/specassay-check/extension.yml
preset specassay 0.5.2 .specify/presets/specassay/preset.yml

Nothing resolved to 0.5.1 or 0.4.12 anywhere in the project, so there was nothing to stop and report and nothing to patch.

The Gate runs in the installed project: it refuses on the fresh project for the documented reason (no registry file), and after one touch PRD.md reports gate.ok=True with OK, registry empty (0 IDs in PRD.md), which is the honest-green line the README promises.

2. The landing, recorded

CHEATSHEET.md gains a v0.5.2 landed section closing the arc v0.5.2 refiled opens, and its front table gains the PR that landed each issue.

Component Issue Generated PR Merged
Preset #4713 #4717 2026-09-24, KSchlobohm
Extension #4711 #4735 2026-09-24, KSchlobohm
Bundle #4715 #4737 2026-09-24, KSchlobohm, ac53c9f

Each submission doc gains its own dated landed line at the head of its filing history. Every date, merger and SHA above was read off the PR pages today, not recalled.

Not one comment was ever posted on any of the three issues. Every verdict arrived as labels and a generated PR. That is recorded in all four files, because a reader watching for a reply would have seen nothing happen while the whole round completed.

The manifest-freeze rule, stated as the thing this round cost and taught: the validator reads the catalogs and bundle.yml on the default branch, not the tag an issue names, so while a submission is open, main is load-bearing. Do not cut a release at a new version, and do not move catalogs/*.json or bundle.yml, except to move them with a refiling. The reverse was tried first, pinning the catalogs while letting main move, and that is precisely what invalidated three open issues.

Checks, at 460a2b8

138 passed                                     tests/ + extensions/specassay-check/tests/
doc-versions: OK (2 warning(s))                0 refusals
SpecAssay Check (Gate 2): OK (96 registry IDs) gate.ok=True

doc-versions refused twice while this was being written, on 0.4.12 in the new receipt section: once for the version the community catalog carried until #4737, once for the versions checked for and not found. Both are provenance, so both now carry <!-- specassay:provenance --> rather than being reworded away. The two warnings are PRD.md:119 and :148, 0.5.0 observations, untouched here and due to refuse at 0.5.3.

Two em dashes were added, on purpose. Both sit inside fenced blocks quoting the CLI's own output (specassay v0.5.2 — SpecAssay, — enabled — priority 10). The receipt is what the tool printed; editing its punctuation would make it a paraphrase. No em dash was added to any prose.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Vzf6Sh5RVMKdFqaXKFF2FY


Generated by Claude Code

All three submissions merged on 2026-09-24, each by KSchlobohm, each as a
pull request the submission workflow generated from the issue: the preset
#4713 by PR #4717, the extension #4711 by PR #4735, the bundle #4715 by
PR #4737 at commit ac53c9f. Not one comment was ever posted on any of the
three issues; every verdict arrived as labels and a generated PR, which is
worth writing down because a reader watching for a reply would have seen
nothing happen while the round completed.

The CHEATSHEET gains a **v0.5.2 landed** section closing the arc that
**v0.5.2 refiled** opens, its front table gains the PR that landed each
issue, and the round's lesson is stated as a rule: the validator reads the
catalogs and bundle.yml on the DEFAULT BRANCH, so freeze the manifest while
a submission is open. The reverse was tried first, pinning the catalogs and
letting main move, and that is what cost this round.

It also gains the bundle install receipt, which could not be taken until
#4737 merged because the community catalog read 0.4.12 until then. Taken in
a clean specify init project outside this repository, Spec Kit 1.0.5,
Python 3.11.15, Linux, on 2026-09-24:

  specify bundle info specassay  -> both components resolve at 0.5.2 from
                                    the community catalog, which refuses to
                                    install from there (discovery-only)
  specify bundle install specassay (README catalog path)
                                 -> Installed 'specassay' (2 added)
  bundle/preset/extension list   -> 0.5.2, 0.5.2, 0.5.2

Confirmed on disk as well as in the listings: bundle-records.json and both
installed manifests read 0.5.2, and nothing resolved to the versions those
entries carried before.

The two em dashes added are inside fenced blocks quoting the CLI's own
output, so they stay: the receipt is what the tool printed.

Checks on this commit: 138 passed, doc-versions OK (2 warnings, both
PRD.md 0.5.0 observations), Gate 2 OK on 96 registry IDs.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Vzf6Sh5RVMKdFqaXKFF2FY
@coderabbitai

coderabbitai Bot commented Sep 24, 2026

Copy link
Copy Markdown

Important

  • 🔍 Trigger review

This repository does not receive automatic reviews because it has fewer than 10 stars.

⚙️ Run configuration

Configuration used: defaults

Review profile: CHILL

Plan: Advanced

Run ID: 91fbd4aa-5c32-452e-aa72-b89c38358fd3


Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

Comment @coderabbitai help to get the list of available commands.

@rdryfoos
rdryfoos merged commit 0dfb8a4 into main Sep 24, 2026
2 checks passed
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.

2 participants