GitHub issue-comment metadata causes false PR-target ambiguity in automatic approval
Maintainers usually reply within 1 day
Nobody has claimed this yet.
Assessment
- Difficulty
- 2/5
- Estimated time
- 1-3 hours
- Newbie friendliness
- 76/100
- Issue type
- Bug
- Clarity
- Clearly specified
- Activity status
- Active
- Tech stack
- json, rust
- Domain
- developer-experience, security
Research direction
Read codex-rs/core/assets/consequential_tool_message_templates.json and compare the add_comment_to_issue entry with the exposed tool contract. Update the target label and approval question to refer to an issue or pull request while preserving the pr_number field, then verify both target types through the production connector and inspect the returned canonical URLs. The public template change alone does not prove the automatic approval decision is fixed.
Written by the indexing model from the issue text.
Description
Summary
The managed GitHub tool github_add_comment_to_issue advertises a PR-only target, although its name, body description and observed behavior include ordinary issues. Automatic approval can therefore approve one ordinary-issue comment and reject a later comment to the same verified issue as an unauthorized target ambiguity.
This report is sanitized. Repository names, issuer names, private report content, account identifiers, credentials and complete conversation logs are excluded.
Exposed tool metadata
Description: "Create a top-level PR Conversation comment (Issue comment)."
Parameters:
- repo_full_name: string
- pr_number: number; description "Pull request number in the repository."
- comment: string; description "Top-level comment body to add to the issue thread."
Reproduction from October 17, 2026 UTC
- Fetch an ordinary issue from a private repository; confirm the issue number and canonical /issues/ URL.
- Add a claim comment with github_add_comment_to_issue using that repository and number in pr_number. It succeeds.
- Later submit a handoff comment through the same tool to the same ordinary issue.
- Automatic approval rejects the request with this rationale, identifiers sanitized:
"This externally visible comment is intended for issue #17, but the tool argument is explicitly pr_number: 17 and the target type is ambiguous, creating a material risk of posting to the wrong PR or failing; the user has not authorized that exact target ambiguity."
Expected behavior
The tool contract and approval presentation should state that the target can be either an issue or a pull request. Once the repository and numbered issue are verified, the misleading legacy argument name should not create a second, fictitious target.
Primary evidence
GitHub documents that issue comments and PR conversation comments share the Issues endpoints:
https://docs.github.com/en/rest/issues/comments#create-an-issue-comment
OpenAI Codex source at ce0323f2f2d3b25e82806cd555237da3db4e1897 also labels this managed GitHub tool's target "Pull request" and asks to comment "on a pull request":
https://github.com/openai/codex/blob/ce0323f2f2d3b25e82806cd555237da3db4e1897/codex-rs/core/assets/consequential_tool_message_templates.json
Relevant entry: tool_title add_comment_to_issue; connector_id connector_76869538009648d5b282a4bb21c3d157; source_tool_index 0.
Smallest correction
- Managed connector description: "Create a top-level conversation comment on a GitHub issue or pull request."
- Keep the existing pr_number wire field for compatibility; describe it as the issue or pull-request number, documenting its legacy name.
- Update the bundled approval label and question to "Issue or pull request" / "an issue or pull request".
- Verify ordinary-issue and PR conversation comments through the same production connector, checking the returned canonical target URL. Preserve normal approval checks.
The accompanying patch changes only the public Codex approval presentation. It does not modify the managed connector schema, and applying it alone must not be claimed to resolve automatic-review decisions.
Separate incident
A branch-create attempt returned "Automatic approval review failed: authorization changed before approval could be applied". Public Codex source maps that message to StaleAuthorization when authorization or review-context revisions change during review. The branch and PR subsequently succeeded. This is distinct from the comment metadata problem; the original internal attempt count and precise changing evidence were not retained, so no specific race is claimed proven.
Source:
https://github.com/openai/codex/blob/ce0323f2f2d3b25e82806cd555237da3db4e1897/codex-rs/core/src/guardian/review_request.rs
https://github.com/openai/codex/blob/ce0323f2f2d3b25e82806cd555237da3db4e1897/codex-rs/ext/guardian-reviewer/src/retry.rs
Public Codex approval-template patch
diff --git a/codex-rs/core/assets/consequential_tool_message_templates.json b/codex-rs/core/assets/consequential_tool_message_templates.json
--- a/codex-rs/core/assets/consequential_tool_message_templates.json
+++ b/codex-rs/core/assets/consequential_tool_message_templates.json
@@ -9,7 +9,7 @@
"template_params": [
{
"name": "pr_number",
- "label": "Pull request"
+ "label": "Issue or pull request"
},
{
"name": "repo_full_name",
@@ -20,7 +20,7 @@
"label": "Comment"
}
],
- "template": "Allow {connector_name} to add a comment to a pull request?"
+ "template": "Allow {connector_name} to add a comment to an issue or pull request?"
},
{
"source_tool_index": 1,
- Dominant language
- Rust
- Stars
- 129k
- Forks
- 20.3k
- Avg merge
- 1m
- Merged PRs (30d)
- 998
Getting set up
- No Dockerfile or Docker Compose file
- No pull request template
- Read the contributing guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
More from openai/codex
-
Python SDK: CodexClient.close() leaves the app-server's stdout/stderr pipes open (and never waits after kill)Possibly taken @akhil9tiet claimed this today. Openapp-server bug
Difficulty 2/5 1-3 hours Newbie friendliness 82/100
openai/codex#54908 · 2 comments ·
Maintainers usually reply within 1 day
-
bug CLI tool-calls windows-os
Difficulty 2/5 1-3 hours Newbie friendliness 84/100
Maintainers usually reply within 1 day
-
Clarification about NODE_REPL_TRUSTED_BROWSER_CLIENT_SHA256S in the VS Code extension environmentOpendocumentation extension windows-os
Difficulty 2/5 1-3 hours Newbie friendliness 64/100
openai/codex#54780 · 1 comment ·
Maintainers usually reply within 1 day
-
bug context
Difficulty 2/5 Under an hour Newbie friendliness 74/100
Maintainers usually reply within 1 day
-
Standalone Unix updater reports success when curl fails (`curl | sh` masks the exit status); Windows path is unaffectedPossibly taken @akhil9tiet claimed this 2 days ago. Openbug CLI windows-os
Difficulty 2/5 1-3 hours Newbie friendliness 72/100
openai/codex#54465 · 1 comment ·
Maintainers usually reply within 1 day
Similar issues
-
Difficulty 2/5 1-3 hours Newbie friendliness 68/100
nautechsystems/nautilus_trader#5365 ·
Maintainers usually reply within 1 day
-
documentation
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
IntersectMBO/developer-experience#347 ·
Maintainers usually reply within 2 days
-
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
registrystack/registry-stack#2168 ·
Maintainers usually reply within 1 day
-
bug crate: 3d help wanted triage
Difficulty 2/5 1-3 hours Newbie friendliness 78/100
kornia/kornia-rs#1203 · 1 comment ·
Maintainers usually reply within 1 day
-
Difficulty 2/5 1-3 hours Newbie friendliness 76/100
Maintainers usually reply within 1 day