🐛 fix(prompt): create the repo as <name> when prompt push uses --owner - by codekiln · Pull Request #824 · codekiln/langstar

codekiln · 2026-10-14T04:13:55Z

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Fixes #820

langstar prompt push --owner - --repo <name> to a repo that does not exist yet sent -/<name> as the new repo's repo_handle, and the API answered 400 "handle must be lowercase alphanumeric, hyphen, or underscore, starting with a-z, or a valid UUID". The CLI printed that as a warning, pushed anyway, and failed with 404 "Repository not found". create_repo in sdk/src/prompts.rs now sends only <name> for a -/<name> handle.

Test

test_cli_push_creates_new_repo_with_owner_dash in cli/tests/prompt_structured_test.rs pushes a structured prompt with --owner - to a repo that does not exist, pulls it back as JSON and asserts on the template and schema title. It deletes the repo afterwards. Without the fix, the push fails with 404 "Repository not found". It pushes with --schema because a push without --schema has its own bug, fixed in 🐛 fix(prompt): send a ChatPromptTemplate manifest when pushing without --schema · Pull Request #823.

LangChain's Python and Go clients send only the repo name

The Python client, langsmith-sdk, splits owner/name with parse_prompt_identifier, accepts an owner of - or the workspace's own handle, and posts only the name as "repo_handle": prompt_name, so the owner never reaches the API.
langsmith-go's RepoNewParams has a repo_handle field and no owner field, so its caller passes a bare name.

Decisions for codekiln

Handles with a named owner are still sent as given

A handle with a named owner, such as acme/name, still goes to the API unchanged. According to its 400 message, the API accepts only a handle of lowercase letters, digits, hyphens and underscores, or a UUID, so it probably rejects acme/name. langsmith-sdk sends only the bare name when the owner is - or the workspace's own handle, so I recommend filing a follow-up issue to make create_repo do the same for the workspace's own handle. I could not test a named owner because this workspace has no handle.

🤖 Generated with Claude Code

…repo #820 🐛 fix(prompt): prompt push --owner - fails when it creates a new repo Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

github-actions · 2026-10-14T04:14:52Z

Unit Test Results

227 tests +2 227 ✅ +2 3s ⏱️ ±0s
1 suites ±0 0 💤 ±0
1 files ±0 0 ❌ ±0

Results for commit 2fd54d1. ± Comparison against base commit ab158b1.

♻️ This comment has been updated with latest results.

Copilot

🟢 Approval recommended

The focused fix satisfies issue #820 and includes appropriate unit and integration coverage.

0 open findings

What changed in this PR

Fixes prompt repository creation when --owner - is used by sending the bare repository name to LangSmith.

Changes:

Normalizes -/name handles during repository creation.
Adds unit and end-to-end integration coverage for create-on-push behavior.

File	Description
`sdk/src/prompts.rs`	Normalizes current-workspace handles and adds unit tests.
`cli/tests/prompt_structured_test.rs`	Verifies push, pull, and cleanup for a newly created repository.

🧠 Review effort: Balanced

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

github-actions · 2026-10-14T04:15:57Z

SDK Integration Test Results

339 tests +2 339 ✅ +2 15s ⏱️ ±0s
14 suites ±0 0 💤 ±0
1 files ±0 0 ❌ ±0

Results for commit 2fd54d1. ± Comparison against base commit ab158b1.

♻️ This comment has been updated with latest results.

github-actions · 2026-10-14T04:16:14Z

Integration Test Results

0 tests - 362 0 ✅ - 359 0s ⏱️ - 4m 30s
0 suites - 15 0 💤 ± 0
0 files - 1 0 ❌ - 3

Results for commit 2fd54d1. ± Comparison against base commit ab158b1.

♻️ This comment has been updated with latest results.

codekiln

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Three findings on the title and body, which have no line to sit on, and one inline.

1. Title. Quoted: "send the repo name, not -/, when push creates a repo". It makes the reader hold -/<name> as an option nobody proposed in order to learn one fact. Rule: My/AI/Rule/State the Positive. Rewrite: 🐛 fix(prompt): create the repo as <name> when prompt push uses --owner -. Use the same subject as the squash-merge subject.

2. Heading. Quoted: "## Decisions for codekiln: leave owner/name creation as it was". Read alone, "owner/name creation" and "as it was" each send the reader into the section to learn what is being left alone. The heading of a decision should be the decision. Rule: My/AI/Rule/How to Communicate Effectively With Me/Use Headings to Serve a Purpose. Rewrite: ## Handles with a named owner are still sent as given.

3. Pointer with no link or title. Quoted: "It pushes with --schema because the plain push is fixed separately in #823; this branch does not depend on that PR." A bare #823 says nothing about the request, and the second clause answers a question nobody asked. Rule: My/AI/Rule/How to Communicate Effectively With Me/A pointer carries a link, an id with a slug, and a reason. Rewrite: It pushes with --schemabecause a push without--schema has its own bug, fixed in [🐛 fix(prompt): send a ChatPromptTemplate manifest when pushing without --schema · Pull Request #823](https://github.com/codekiln/langstar/pull/823).

Checked and found accurate: the three pinned links and line anchors in the body (parse_prompt_identifier at utils.py L711-L743, "repo_handle": prompt_name at client.py L9904, RepoNewParams at repo.go L310-L320) land on the code the body describes.

#820 🐛 fix(prompt): prompt push --owner - fails when it creates a new repo Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

codekiln · 2026-10-14T04:21:58Z

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

AI (codekiln-review): Edited the title and body for the three findings.

Title: "send the repo name, not -/, when push creates a repo" is now "🐛 fix(prompt): create the repo as when prompt push uses --owner -".
Heading: "## Decisions for codekiln: leave owner/name creation as it was" is now "## Decisions for codekiln" with "### Handles with a named owner are still sent as given" as its first line. I kept the "Decisions for codekiln" parent so a search for that label still finds the section.
Pointer: "It pushes with --schema because the plain push is fixed separately in 🐛 fix(prompt): send a ChatPromptTemplate manifest when pushing without --schema #823; this branch does not depend on that PR." is now "It pushes with --schema because a push without --schema has its own bug, fixed in 🐛 fix(prompt): send a ChatPromptTemplate manifest when pushing without --schema · Pull Request #823."
Added to the decision: "I could not test a named owner because the workspace has no handle; upstream accepts only - or the tenant's own handle and sends only the bare name." It replaces "but I did not test a named owner."

Copilot

🟢 Approval recommended

The focused fix satisfies issue #820 and includes appropriate unit and integration coverage.

0 open findings

🧠 Review effort: Balanced

codekiln

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Two findings on the body, which has no line to sit on. Everything else read clean: the title, the doc comments, the test and the earlier rounds' rewrites. The three pinned upstream links and their line anchors land on the code the body describes.

1. Decision paragraph. Quoted: "This change strips only a leading -/. A named owner (acme/name) is still sent as given, and the API's error message suggests it rejects a slash too. Upstream drops every owner, so sending the bare name for any owner is probably right. I could not test a named owner because the workspace has no handle; upstream accepts only - or the tenant's own handle and sends only the bare name. Recommendation: a follow-up issue." Cost: the reader meets "upstream" twice saying the same thing, has to work out what "it" and "a follow-up issue" are (no link, no title, no statement of what the issue would change), and the reason for not testing is followed by a clause that has nothing to do with testing. The heading above it already says the first sentence. Rule: My/AI/Rule/How to Communicate Effectively With Me/Never make the reader resolve a reference and My/AI/Rule/Prune useless commandments. Rewrite, to replace the whole paragraph: A handle with a named owner, such as acme/name, still goes to the API unchanged, and the API's error message lists no slash among the allowed characters, so the API probably rejects it. langsmith-sdk sends only the bare name whatever the owner, so I recommend a follow-up issue that does the same here. I could not test a named owner because this workspace has no handle.

2. Heading and first bullet of the upstream section. Quoted: "## How langsmith-sdk and langsmith-go send the handle" and "- langsmith-sdk splits the identifier with parse_prompt_identifier, accepts an owner of - or the tenant's own handle, and posts "repo_handle": prompt_name, so the owner never reaches the API." Read alone, the heading names two packages the reader has not been told are LangChain's Python and Go clients, and "the handle" could be any handle; the bullet repeats the unintroduced name, then "the identifier" and "the tenant's own handle". The point of the section is that both clients send only the repo name. Rule: My/AI/Rule/How to Communicate Effectively With Me/Use Headings to Serve a Purpose and My/Pref/Writing/Be specific and explicit. Rewrite: heading ## LangChain's Python and Go clients send only the repo name, first bullet The Python client, langsmith-sdk, splits owner/name with [parse_prompt_identifier](<the existing pinned link>), accepts an owner of - or the workspace's own handle, and posts only the name as ["repo_handle": prompt_name](<the existing pinned link>), so the owner never reaches the API.

codekiln · 2026-10-14T04:24:44Z

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

AI (codekiln-review): I rewrote the two body passages the review named. I checked the facts first: the API's 400 message lists lowercase alphanumeric, hyphen, underscore and a UUID, with no slash. parse_prompt_identifier returns the owner and name separately, and create_prompt posts only prompt_name as repo_handle. langsmith-go's RepoNewParams has no owner field.

Decision paragraph. Replaced "This change strips only a leading -/. A named owner (acme/name) is still sent as given, and the API's error message suggests it rejects a slash too. Upstream drops every owner, so sending the bare name for any owner is probably right. I could not test a named owner because the workspace has no handle; upstream accepts only - or the tenant's own handle and sends only the bare name. Recommendation: a follow-up issue." with "A handle with a named owner, such as acme/name, still goes to the API unchanged, and the API's error message lists no slash among the allowed characters, so the API probably rejects it. langsmith-sdk sends only the bare name whatever the owner, so I recommend a follow-up issue that does the same here. I could not test a named owner because this workspace has no handle."
Upstream section. Heading "## How langsmith-sdk and langsmith-go send the handle" is now "## LangChain's Python and Go clients send only the repo name". First bullet now reads "The Python client, langsmith-sdk, splits owner/name with parse_prompt_identifier, accepts an owner of - or the workspace's own handle, and posts only the name as "repo_handle": prompt_name, so the owner never reaches the API." The two pinned links are unchanged.

codekiln

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

One finding on the body, which has no line to sit on. The title, the code, the doc comments, the test, the earlier rounds' rewrites and the three pinned upstream links with their line anchors read clean.

1. Decision paragraph. Quoted: "langsmith-sdk sends only the bare name whatever the owner, so I recommend a follow-up issue that does the same here." Cost: an issue cannot send a name, so the reader works out that the subject should be a change to create_repo; and "whatever the owner" is broader than the code the body pins, because create_prompt calls _current_tenant_is_owner and raises an error for any owner other than - or the workspace's own handle, which the section above states itself. The first sentence of the paragraph also lists what the error message leaves out ("lists no slash") instead of what it allows. Rules: My/Pref/Writing/Give every sentence a subject that can do its verb, My/Pref/Writing/Be specific and explicit and My/AI/Rule/State the Positive. Rewrite of the paragraph: A handle with a named owner, such as acme/name, still goes to the API unchanged. The API's error message allows only lowercase letters, digits, hyphens, underscores or a UUID, so it probably rejects that handle. langsmith-sdk sends only the bare name when the owner is -or the workspace's own handle, so I recommend filing a follow-up issue to makecreate_repo do the same for the workspace's own handle. I could not test a named owner because this workspace has no handle.

codekiln · 2026-10-14T04:27:16Z

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

AI (codekiln-review): I rewrote the Decisions paragraph as the review asked. I checked it against _current_tenant_is_owner in langsmith-sdk at 9966fd69 (return owner == "-" or settings.tenant_handle == owner) and against the API message quoted in the body, and both match. The sentence that changed:

langsmith-sdk sends only the bare name when the owner is - or the workspace's own handle, so I recommend filing a follow-up issue to make create_repo do the same for the workspace's own handle.

The paragraph's first two sentences now say what the API allows ("lowercase letters, digits, hyphens, underscores or a UUID") instead of what it leaves out.

codekiln

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Two findings on the body, which has no line to sit on. The title, the code, the doc comments, the test, the earlier rounds' rewrites, the pinned upstream links and their line anchors (checked against the pinned commits with gh api) and the claims about _current_tenant_is_owner and the bare repo_handle read clean.

1. Decision paragraph, second sentence. Quoted: "The API's error message allows only lowercase letters, digits, hyphens, underscores or a UUID, so it probably rejects that handle." Cost: an error message cannot allow anything, so the reader works out that the API allows and the message only reports it, and "that handle" sends them back one sentence. Rule: My/Pref/Writing/Give every sentence a subject that can do its verb. Rewrite: According to its 400 message, the API accepts only a handle of lowercase letters, digits, hyphens and underscores, or a UUID, so it probably rejects acme/name.

2. Test section, fourth sentence. Quoted: "Without the fix in this pull request, it fails with the 404 above." Cost: "the 404 above" sends the reader back past a heading to the first paragraph to learn which 404. Rule: My/AI/Rule/How to Communicate Effectively With Me/Never make the reader resolve a reference. Rewrite: Without the fix, the push fails with 404 "Repository not found".

codekiln · 2026-10-14T04:29:43Z

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

AI (codekiln-review): Both findings on the body are fixed.

Decisions paragraph, second sentence. Was: "The API's error message allows only lowercase letters, digits, hyphens, underscores or a UUID, so it probably rejects that handle." Now: "According to its 400 message, the API accepts only a handle of lowercase letters, digits, hyphens and underscores, or a UUID, so it probably rejects acme/name."
Test section, fourth sentence. Was: "Without the fix in this pull request, it fails with the 404 above." Now: "Without the fix, the push fails with 404 "Repository not found"."

The 404 text and the 400 wording match what the first paragraph of the body already states.

Copilot

🟢 Approval recommended

The focused normalization correctly addresses issue #820 and is covered by unit and integration tests.

0 open findings

🧠 Review effort: Balanced

Copilot

🟢 Approval recommended

The narrow fix addresses the reported failure with regression coverage and no identified blocking issues; tests were not executed during review.

0 open findings

🧠 Review effort: Balanced

Copilot

🟢 Approval recommended

The narrow fix has regression coverage and no identified blocking issues; tests were not executed during review.

0 open findings

🧠 Review effort: Balanced

Copilot

🟢 Approval recommended

The localized fix addresses the reported failure with appropriate regression coverage and no identified blocking issues; tests were not executed.

0 open findings

🧠 Review effort: Balanced

Copilot

🟡 Changes recommended

Both command setups in the new test reference an undeclared helper, preventing the test target from compiling.

1 open issue

Use declared config_dir helper instead of missing home module · New

_{Review effort: Balanced}

+    push_cmd.env("HOME", home::empty_home());
+    push_cmd.args([
+        "prompt",
+        "push",
+        "--owner",
+        "-",
+        "--repo",
+        &repo_name,
+        "--template",
+        "Owner dash push: {input}",
+        "--input-variables",
+        "input",
+        "--schema",
+        schema_path,
+    ]);
+    push_cmd.assert().success();
+
+    // VERIFY: the pulled manifest has the pushed template and schema
+    let mut pull_cmd = Command::new(&bin);
+    pull_cmd.env("HOME", home::empty_home());


🐛 fix(prompt): send the repo name, not -/<name>, when push creates a …

1559d2c

…repo #820 🐛 fix(prompt): prompt push --owner - fails when it creates a new repo Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

Copilot AI balanced review requested due to automatic review settings October 14, 2026 04:13

codekiln added this to the reboot milestone Oct 14, 2026

codekiln added bug Something isn't working wip work in progress labels Oct 14, 2026

Copilot started reviewing on behalf of codekiln October 14, 2026 04:14

View session

Copilot AI reviewed Oct 14, 2026

View reviewed changes

codekiln commented Oct 14, 2026

View reviewed changes

Comment thread cli/tests/prompt_structured_test.rs Outdated

📚 docs(prompt): say what the owner-dash test does

f9a6ab4

#820 🐛 fix(prompt): prompt push --owner - fails when it creates a new repo Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

Copilot AI balanced review requested due to automatic review settings October 14, 2026 04:21

Copilot started reviewing on behalf of codekiln October 14, 2026 04:21

View session

codekiln changed the title ~~🐛 fix(prompt): send the repo name, not -/<name>, when push creates a repo~~ 🐛 fix(prompt): create the repo as <name> when prompt push uses --owner - Oct 14, 2026

Copilot AI reviewed Oct 14, 2026

View reviewed changes

codekiln commented Oct 14, 2026

View reviewed changes

codekiln added ready-for-review Green CI, no open threads, Copilot settled; ready for codekiln to review and merge and removed wip work in progress labels Oct 14, 2026

Merge branch 'main' into m19-i820-prompt-push-owner-dash

9ce26e9

Copilot AI balanced review requested due to automatic review settings October 14, 2026 14:20

codekiln removed the ready-for-review Green CI, no open threads, Copilot settled; ready for codekiln to review and merge label Oct 14, 2026

codekiln added the wip work in progress label Oct 14, 2026

Copilot started reviewing on behalf of codekiln October 14, 2026 14:22

View session

Copilot AI reviewed Oct 14, 2026

View reviewed changes

codekiln added the priority: medium Review after high: bug fixes and features with no open decision label Oct 14, 2026

Merge branch 'main' into m19-i820-prompt-push-owner-dash

6141cc8

Copilot AI balanced review requested due to automatic review settings October 14, 2026 19:52

Copilot started reviewing on behalf of codekiln October 14, 2026 19:52

View session

Copilot AI reviewed Oct 14, 2026

View reviewed changes

codekiln requested a balanced review from Copilot October 14, 2026 19:57

Copilot started reviewing on behalf of codekiln October 14, 2026 19:58

View session

Copilot AI reviewed Oct 14, 2026

View reviewed changes

codekiln requested a balanced review from Copilot October 14, 2026 20:00

Copilot started reviewing on behalf of codekiln October 14, 2026 20:00

View session

Copilot AI reviewed Oct 14, 2026

View reviewed changes

codekiln added ready-for-review Green CI, no open threads, Copilot settled; ready for codekiln to review and merge and removed wip work in progress labels Oct 14, 2026

Merge branch 'main' into m19-i820-prompt-push-owner-dash

2fd54d1

Copilot AI balanced review requested due to automatic review settings October 27, 2026 21:08

Copilot started reviewing on behalf of codekiln October 27, 2026 21:09

View session

Copilot AI reviewed Oct 27, 2026

View reviewed changes

codekiln removed the ready-for-review Green CI, no open threads, Copilot settled; ready for codekiln to review and merge label Oct 27, 2026

codekiln added the wip work in progress label Oct 27, 2026 — with Claude

Conversation

codekiln commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Test

LangChain's Python and Go clients send only the repo name

Decisions for codekiln

Handles with a named owner are still sent as given

Uh oh!

github-actions Bot commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Unit Test Results

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

github-actions Bot commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

SDK Integration Test Results

Uh oh!

github-actions Bot commented Oct 14, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Integration Test Results

Uh oh!

codekiln left a comment

Choose a reason for hiding this comment

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Uh oh!

Uh oh!

codekiln commented Oct 14, 2026

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

codekiln left a comment

Choose a reason for hiding this comment

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Uh oh!

codekiln commented Oct 14, 2026

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Uh oh!

codekiln left a comment

Choose a reason for hiding this comment

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Uh oh!

codekiln commented Oct 14, 2026

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Uh oh!

codekiln left a comment

Choose a reason for hiding this comment

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Uh oh!

codekiln commented Oct 14, 2026

from AI(claude-code, claude-sonnet-5-5, langstar i809-dataset-export-all)

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟢 Approval recommended

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

🟡 Changes recommended

Uh oh!

Reviewers

Uh oh!

codekiln commented Oct 14, 2026 •

edited

Loading

github-actions Bot commented Oct 14, 2026 •

edited

Loading

github-actions Bot commented Oct 14, 2026 •

edited

Loading

github-actions Bot commented Oct 14, 2026 •

edited

Loading