Recommendations · Jinear (demo) · as of 27 Aug 2026
9 open.
save_issuev1Agents are failing to set issue priority with strings.
What to change
Update the `priority` argument to accept string values like 'High' or 'Low' in addition to integers, or update its description to clarify that only integer values are supported.
The primary failure mode is `arg:schema_violation` because agents are attempting to pass string names for priority instead of the required integer.
Impact
Failure mix
list_issues, get_issue, list_issue_statuses, get_userNo earlier recorded version to compare.confidence: highlist_issuesv1Tool is unreliable due to frequent upstream dependency failures.
Heads-up
Investigate the root cause of the high rate of `dep:upstream_unreachable` and `dep:rate_limited` errors with the backend service.
The primary failure mode is backend unavailability, leading to a high unresolved task rate and user-facing errors.
Impact
Failure mix
save_issue, list_projects, list_issue_statuses, get_issueNo earlier recorded version to compare.confidence: highget_issuev1High rate of backend failures and entity not found errors.
Heads-up
Investigate the high rate of 'dep:upstream_unreachable' and 'resource:not_found' errors; the former suggests a backend health issue, while the latter may indicate the model is hallucinating issue IDs.
A significant portion of calls fail due to backend timeouts or because the requested issue does not exist, leading to a high rate of unresolved tasks and tool abandonment.
Impact
Failure mix
list_issues, list_usersNo earlier recorded version to compare.confidence: highlist_issue_labelsv1High rate of backend dependency failures causing unresolved tasks.
Heads-up
Investigate the high incidence of `dep:upstream_unreachable` and `dep:rate_limited` errors with the backend service.
The primary failure mode is backend dependency errors, which the tool owner cannot fix via the contract and require upstream investigation.
Impact
Failure mix
list_teams, list_project_labels, save_issueNo earlier recorded version to compare.confidence: highget_projectv1High rate of backend timeouts when retrieving projects.
Heads-up
Investigate the high rate of 'database connection timeout' errors with the backend service responsible for project lookups.
A significant portion of calls are failing due to upstream dependency issues, making the tool unreliable.
Impact
Failure mix
list_teams, list_projectsNo earlier recorded version to compare.confidence: highget_documentv1Tool fails frequently when users guess document IDs/slugs.
What to change
Add a search capability to the tool, or guide the model to use `list_documents` first to find a valid ID before calling `get_document`.
The tool is almost always called with non-existent IDs, leading to a high rate of unresolved tasks and abandonment.
Impact
Failure mix
list_documentsNo earlier recorded version to compare.confidence: highlist_projectsv1Tool is failing frequently due to upstream dependency issues.
Heads-up
Investigate the high rate of `dep:upstream_unreachable` and `dep:rate_limited` errors with the backend service that serves project lists.
The primary driver of failure is unresolved tasks, caused almost entirely by backend dependency errors that the tool owner cannot fix via the contract.
Impact
Failure mix
list_teams, get_project, list_issues, save_issueNo earlier recorded version to compare.confidence: highlist_teamsv1High rate of backend failures causing unresolved tasks.
Heads-up
Investigate the high rate of `dep:upstream_unreachable` and `dep:rate_limited` errors with the backend service owner.
The primary failure mode is backend dependency errors, which are outside the tool contract owner's direct control but are severely impacting user success.
Impact
Failure mix
list_projects, list_project_labels, list_issues, save_issueNo earlier recorded version to compare.confidence: highget_diffv1Tool is mocked but failing with backend dependency errors.
Heads-up
Investigate why a mocked tool is returning `dep:upstream_unreachable` errors like 'internal server error' and 'database connection timeout'.
The tool is marked as mocked, but its primary failure mode is unreachable backends, which should not happen.
Impact
Failure mix