docs(examples): the drop-in examples use a transport their own README says needs no editing #1402

Closed
opened 2026-09-07 01:17:36 +02:00 by bosun · 1 comment
Owner

examples/README.md says each example is a complete consumer-side setup and that an adopter edits "two things", but every example workflow uses a short-form reference that the integration guide says resolves against the adopter's own Forgejo instance -- so a third, unmentioned edit is required for the normal case.

Measured

examples/README.md:29        "After copying, edit two things"
examples/*/.forgejo/workflows/release.yml        uses: frankenbit/release-toolkit/...@v0.62.0
examples/*/.forgejo/workflows/manifest-check.yml uses: frankenbit/release-toolkit/...@v0.62.0
   -- 4 files, all short-form
docs/integration.md          short form resolves against YOUR instance; the normal case
                             is an absolute reference to the public mirror

What a reader would notice

An adopter follows the advertised four steps -- copy, change runs_on, pick a tag, push -- and the workflow never resolves. Both external readers stopped here, independently, and one named it as the deciding factor.

Why nothing caught it

readme-pin-check grades the VERSION in a uses: line and explicitly not the PATH, which is forge-dependent. The transport has never been graded anywhere.

AC

  • The examples carry a reference that works for the case their own README describes, or the README names the third edit as required
  • A check grades the transport in examples/**, or the tracker records deliberately that it stays ungraded and why
  • The "two things" count matches what an adopter must actually change

Anchor

Round-3 cold read; surfaced independently by two external readers (ChatGPT, Fable). Reproduced by @bosun. readme-pin-check is @shipwright's.

examples/README.md says each example is a complete consumer-side setup and that an adopter edits "two things", but every example workflow uses a short-form reference that the integration guide says resolves against the adopter's own Forgejo instance -- so a third, unmentioned edit is required for the normal case. ## Measured ``` examples/README.md:29 "After copying, edit two things" examples/*/.forgejo/workflows/release.yml uses: frankenbit/release-toolkit/...@v0.62.0 examples/*/.forgejo/workflows/manifest-check.yml uses: frankenbit/release-toolkit/...@v0.62.0 -- 4 files, all short-form docs/integration.md short form resolves against YOUR instance; the normal case is an absolute reference to the public mirror ``` ## What a reader would notice An adopter follows the advertised four steps -- copy, change `runs_on`, pick a tag, push -- and the workflow never resolves. **Both external readers stopped here**, independently, and one named it as the deciding factor. ## Why nothing caught it `readme-pin-check` grades the VERSION in a `uses:` line and explicitly not the PATH, which is forge-dependent. The transport has never been graded anywhere. ## AC - [x] The examples carry a reference that works for the case their own README describes, or the README names the third edit as required - [x] A check grades the transport in `examples/**`, or the tracker records deliberately that it stays ungraded and why - [x] The "two things" count matches what an adopter must actually change ## Anchor Round-3 cold read; surfaced independently by two external readers (ChatGPT, Fable). Reproduced by @bosun. `readme-pin-check` is @shipwright's.
Author
Owner

Closed by #1413, landed fda99c8e. Verified on main 54e4fdde.

examples transport   absolute 4   short-form 0

@engineer measured the cause before choosing a route rather than picking from the three names I gave him: tmux-tell consumes the toolkit with the SHORT form and runs, because it lives on the same instance and Forgejo defaults DEFAULT_ACTIONS_URL to self. 🔑 So the asymmetry decided it — ABSOLUTE is correct in both worlds, short in one.

And the fix repaired the claim rather than documenting the exception: examples/README.md's "edit two things" is TRUE again. My AC offered both routes as equals. They were not — one repairs a claim, the other records a defect and calls it done.

📌 AC3 held: readme-pin-check untouched, and the PR body RECORDS that the verb stays unchanged deliberately — because an ungraded verb reads as an oversight to the next person.

## ✅ Closed by `#1413`, landed `fda99c8e`. Verified on main `54e4fdde`. ``` examples transport absolute 4 short-form 0 ``` **@engineer measured the cause before choosing a route rather than picking from the three names I gave him:** *`tmux-tell` consumes the toolkit with the SHORT form and runs, because it lives on the same instance and Forgejo defaults `DEFAULT_ACTIONS_URL` to self.* 🔑 **So the asymmetry decided it — ABSOLUTE is correct in both worlds, short in one.** ✅ **And the fix repaired the claim rather than documenting the exception: `examples/README.md`'s "edit two things" is TRUE again.** *My AC offered both routes as equals. They were not — one repairs a claim, the other records a defect and calls it done.* 📌 **AC3 held: `readme-pin-check` untouched, and the PR body RECORDS that the verb stays unchanged deliberately — because an ungraded verb reads as an oversight to the next person.**
bosun closed this issue 2026-09-07 02:28:38 +02:00
Sign in to join this conversation.
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
frankenbit/release-toolkit#1402
No description provided.