Skip to content

fix(integrations): validate github, gitlab, linear, slack, pagerduty, incident.io, sentry and datadog - #10071

Merged
waleedlatif1 merged 8 commits into
stagingfrom
sweep2/v1
Nov 16, 2026
Merged

waleedlatif1 merged 8 commits into
stagingfrom
sweep2/v1

Conversation

@waleedlatif1

@waleedlatif1 waleedlatif1 commented Nov 16, 2026 •

Copy link
Copy Markdown
Collaborator

Summary

Validates the GitHub, GitLab, Linear, Slack, PagerDuty, incident.io, Sentry and Datadog integrations against each vendor's official API spec, fixes what was wrong, and moves every block mapper in scope onto mapperOperation (added in #10070) so the operation is read from the tool-call context.

  • Every tool's endpoint, method, params, body, response parsing, outputs and auth/scopes was checked against the vendor's machine-readable spec. Claims that could not be verified were left unchanged.
  • Block mappers no longer read params.operation, which is undefined when a block runs as an Agent tool, in Pi or in Human-in-the-loop. Each operation's branch now runs on those paths too. Every newly live branch keeps the model's own args (via toolCallArg), accepts already-typed values, and never overwrites an arg with undefined.
  • Canvas requests for saved workflows are unchanged, except for the bug fixes listed below. No subBlock, operation, tool, param or output id was renamed or removed, and no OAuth scope was added.
  • The check:mapper-operation baseline shrinks for every block fixed here, and generated tool metadata and docs are regenerated.
  • Slack: closing a group DM needs the mpim:write scope, which the Sim Slack app does not request. Adding it would make every connected Slack user reconnect, so it is left for a product decision.

Release notes

GitHub

  • Check if starred now returns starred=false for a repository you have not starred, instead of failing with "Not Found". It now queries GitHub's GraphQL API (viewerHasStarred).
  • List workflow runs with Status Filter "All" (the default) now returns runs. Before, it sent status=all and GitHub answered with zero runs.
  • Get project now works from the canvas. The project number is sent as an integer, which GitHub GraphQL requires; it was rejected every time before.
  • Update project: the Public field now applies. Every run that set it used to fail. A saved workflow that already stores Public = Private (false) will now make the project private. New blocks default to "No change".
  • Trigger workflow: Workflow Inputs JSON is now sent as an object. Canvas runs with inputs used to be rejected by GitHub. Invalid JSON now gives a clear error.
  • Create pull request Draft, Create release Draft/Prerelease, and Fork repository Default Branch Only are now sent as JSON booleans instead of the strings true/false.
  • Update pull request and Update issue: the State dropdown has a new "No change" first option, and an empty state is no longer sent. Saved workflows that store Open or Closed are unchanged.
  • GitHub blocks used as Agent tools (or in Pi / Human-in-the-loop) now honor stored fields with block-specific names: Reaction, Milestone Title/Description/State/Sort, Fork Name, Fork Sort, and Gist Public.
  • Repository Owner/Name no longer show or need a value for project, gist, and search operations, which never used them.
  • Page Number is now available for every list and search operation, and Results Per Page for List branches and Get PR files (advanced).
  • Reviewer Usernames now shows in basic mode for Request PR reviewers.
  • File operations now encode file paths and refs, so names containing #, ?, or + reach the right file and ref.
  • Get latest commit (v2/v3) now fills commit author/committer email and date, commit url/tree/verification/comment_count, and the GitHub account fields on author/committer.
  • Search code v2 now returns text_matches.
  • Get PR details stops fetching file pages when the run is cancelled, and fails with a size error above 100 MiB of file data.

GitLab

  • GitLab tools used as Agent tools (or in Pi / Human-in-the-loop) now apply the field renames and type conversions they were missing. Saved values in fields such as Compare From/To, Release Name/Milestones, Internal Note, SAML Provider, Pipeline Variables/Inputs, Create/Update User fields, Search Users query, Update Member access level, listing filters/sort, and the merge request tri-state options now reach GitLab. Access-level names like "Maintainer" are converted to the integer, and arrays the model sends for assignee IDs or milestones no longer fail.
  • Project and group IDs that are already URL-encoded (for example group%2Fproject, the form GitLab's docs show) now work on every project and SAML-link operation, instead of returning 404 Project Not Found.
  • Issue, merge request, pipeline, job, and user ID path parameters are now validated. An ID containing a path separator or a dot segment fails with a clear error instead of sending a request to a different endpoint.
  • An email, SAML group name, identity provider, or branch name of exactly . or .. is now rejected before the request is sent. Before, it could rewrite a DELETE to hit the parent resource.
  • Fixed a .trim is not a function error that failed the GitLab block when a numeric project/group ID was resolved from a reference.
  • Merge Merge Request has a new optional advanced Commit SHA field (shared with Approve), sent as sha. Merges now succeed on groups and instances that require a SHA. A SHA saved earlier on Approve in the same block now also applies when that block is switched to Merge.
  • New optional fields: Title for Update Issue and Update Merge Request; Due Date and Confidential on Create Issue; Draft on Create Merge Request; Merge When Checks Pass (auto-merge) and Squash Commit Message on Merge Merge Request.
  • Update Invitation used as an Agent tool: when the model sends an empty expiresAt, the invitation's expiry is cleared again. Before, the empty value was dropped, or the call failed when no other field was set. Canvas behavior is unchanged.

Linear

  • Create Customer Tier, Update Customer Tier, and Update Customer Status now work. They failed on every call with a GraphQL field error.
  • Archive Label now works. It retires the label, which stays on issues that already use it but can't be newly assigned. It failed on every call before.
  • Delete Project Status is now "Archive Project Status" and archives the status, since Linear has no delete. It failed on every call before. Linear requires that the status has no active projects and is not the last status of its type.
  • Mark notification read/unread (Update Notification) now works. It failed on every call. A blank readAt from an agent now marks the notification read.
  • Create Issue Relation with "Blocked by" now works: it creates "the related issue blocks this issue". It failed on every call before.
  • Create Customer now sends External IDs, Logo URL, Owner, Annual Revenue, and Organization Size. These were silently dropped before.
  • Update Project: setting Project State always failed because Linear removed that input. The block now has an optional Status ID field (project status ID) for moving the project to a status, and a state without a status ID fails with a clear message. Saved Update Project blocks behave as before: a Status ID entered earlier under a project-status operation is not sent.
  • Create, Get, List, and Update Project now also output status.
  • List Projects with a team now filters on Linear's side, so pages and cursors are correct. Before, a page could come back empty while hasNextPage was true.
  • Read Issues Label IDs (comma-separated) now filters. It was silently ignored before.
  • Update Issue and Update Customer Request: new blocks default Priority/Urgency to "No change" instead of silently resetting the value to 0. Saved blocks keep their stored value.
  • Create Customer Request now requires Link to Issue, because Linear requires an issue for API-created requests. A project ID is no longer sent, since Linear rejects it.
  • Create Workflow State now requires a color, because Linear rejects a state without one.
  • List Customers and List Customer Requests now have an Include Archived switch.
  • Create Cycle is hidden from the operation list because Linear no longer allows creating cycles via its API. Saved blocks keep the operation and get Linear's error as before.
  • The deprecated "Paused" project status type is hidden for new selections. Status Color is no longer required for Create Project Label, Title for Update Issue, or Customer Name for Update Customer.
  • Linear used as an Agent tool (or in Pi / Human-in-the-loop) now applies the block's field mappings and coercions on tool calls and keeps the model's own arguments. Before, values filled in a tool row for renamed fields (for example Customer ID or Project Label ID) were ignored.

Slack

  • Slack v2 tools now work as Agent tools (and in Pi / Human-in-the-loop). Workflow operations that use a custom bot or Slack credential (bookmarks, user groups, pins, canvas access, DND, team info, files, presence, unfurls, and more) used to fail with "Select a Slack credential". Agent session operations used to fail with "Slack authentication token is required".
  • Slack tool calls now apply the values pre-filled in the tool row (thread timestamps, canvas IDs, section criteria, invite users, limits, workflow fields such as channel filters). Model arguments are no longer overwritten by empty fields or defaults: get_team_profile visibility, share_canvas access level, set_user_presence presence, list limits, include/inclusive booleans, and edit_canvas operation.
  • Get Channel History and Get Thread Replies now fetch at most 50 pages per run, even when Max Pages is set higher, and each page is capped at 10 MiB. Saved workflows with Max Pages above 50 now stop at 50 pages. When there is more, the result returns hasMore=true and a nextCursor to continue from. A single Slack page larger than 10 MB now fails instead of loading.
  • Set Agent Session Status and Rename Agent Session no longer require a thread timestamp; leave it empty for session channels.
  • Open View and Push View no longer require a Trigger ID. An Interactivity Pointer alone is accepted, matching Slack's views.open API. A run with neither is rejected before the request is sent.
  • Slack ID-list inputs (DND team info users, user group channels and members, open conversation users, canvas share/revoke IDs) accept comma-separated IDs (U1,U2) as well as JSON arrays. Slack's documented limits now fail early with a clear error: 50 DND users, 1 ID for canvas revoke, 20 IDs for canvas share.

PagerDuty

  • Update Incident no longer fails when a Resolution note saved earlier is still set while the status is Acknowledged or Triggered. The note is now sent only when resolving.
  • Incident and service ID fields now reject ., .., and empty values, and encode other characters. Before, a value like ../users could send the request to a different API path.
  • List operations have a new optional advanced Include Total Count switch that fills the total output, which was always null before.
  • Send Event (trigger) has new optional Timestamp and Custom Details (JSON object) fields.

incident.io

  • List Actions and List Follow-ups now use incident.io's v3 endpoints, because v2 is removed on 31 December 2026. With Page Size and After left blank, they follow pages automatically and return up to 2,500 records, oldest first. If more remain, pagination_meta.after is set. Setting Page Size or After fetches a single page of that size starting at that cursor, with no auto-pagination.
  • List Actions and List Follow-ups now apply the advanced Page Size and After fields, which they ignored before. If a saved block still holds a Page Size or After value left over from another list operation, that value is now sent; clear both fields to get the full list again.
  • List/Show Actions and Follow-ups now fill in creator (id/name/email plus the user/api_key/workflow/alert actor). Before, it was always empty. Follow-ups also return assignee_team and category.
  • Workflow outputs now return the real trigger name (for example incident.updated) instead of "[object Object]".
  • Update Incident no longer sends an incident type, which incident.io never applied. The Incident Type field now appears only on Create Incident.
  • Show Incident now returns select and catalog custom field values (value_option, value_catalog_entry). Before, these came back as empty objects.
  • Create Incident no longer requires a severity.
  • Workflow create/update accepts steps, condition groups, expressions, once_for, incident modes, and delay as JSON arrays or objects (for example from another block's output) as well as JSON strings.
  • New blocks: Alert Status defaults to Any on List Alerts (before, it forced "firing"), and Incident Mode defaults to standard and retrospective.
  • Update Schedule and Update Workflow now offer "No change" for Timezone and State, and leave the field unchanged by default. Before, they forced "UTC" and "active". Saved workflows keep their stored values. Create Schedule still defaults to UTC, Create Workflow to active, and Create Alert Event to firing.
  • Create Workflow called as an Agent tool without a state is still created as a draft. Only a block newly added on the canvas defaults to Active.
  • As an Agent tool (or in Pi / Human-in-the-loop): on Update Action, Update Follow-up, List Alerts, and Create Alert Event, the status the model chooses now wins over the block's status dropdown. Fields filled on the tool row (user email, alert title, follow-up title, action description, and similar) are now applied on tool calls.
  • Create Escalation has new optional Description and Incident ID fields, and its User IDs input accepts arrays.

Sentry

  • List operations (issues, projects, events, releases, teams) now return top-level nextCursor and hasMore, so the block's existing Pagination Cursor / Has More outputs work. Before, they were always empty. metadata.nextCursor still works.
  • Create Release: malformed or non-array Commits JSON now fails with a clear error instead of creating the release without commits. Commits and Projects also accept arrays from block references or agents.
  • Path IDs and slugs containing / or exactly ./.. are now rejected before the request is sent (Create Deploy versions still percent-encode /). Surrounding whitespace in slugs and IDs is trimmed.
  • Update Issue has a new optional New Priority field (low/medium/high).
  • The project filter on List Issues and List Releases is relabelled "Project ID or Slug" (Sentry accepts both). The Events limit placeholder is now 10, matching Sentry's 10-per-page cap for full events.

Datadog

  • As an Agent tool (or in Pi / Human-in-the-loop), the block's field mapping and value coercion now run on tool calls. Non-numeric numbers from the model are dropped instead of being sent to Datadog, and switch values sent as strings (true/false) are read correctly. The model's own args are kept whenever the matching block field is blank. A model's customerImpacted=true on Create Incident is honored.
  • Create Monitor and Create Event: tags supplied as an array (from a model or an upstream block output) are sent instead of failing with a TypeError. A tags value that is only whitespace is now omitted instead of being sent as an empty array.
  • Submit Metrics: a numeric series type (0-3, Datadog's MetricIntakeType) is sent instead of being silently dropped.
  • Update Incident: an incident ID with surrounding whitespace is now trimmed in the request body as well as in the URL.
  • New outputs: Create Event event.id_str (exact event ID), Get Incident included (resources requested via include), and Get SLO History errors (partial query errors returned with a 200).
  • Create Event's description now says it uses Datadog's deprecated v1 events endpoint, which is available only to orgs that used it before.

Review follow-ups

  • Slack workflow block: comma-separated ID lists (for example U1,U2) pass through to the tool instead of failing in the block.
  • GitHub file paths reject .. segments; GitLab rejects resource ids of exactly . or ...
  • Sentry: a commits value that is neither a list nor a string now errors instead of being dropped.
  • Linear, GitHub and incident.io outputs that the vendor returns as null are marked nullable.

Testing

  • Each integration (or batch of blocks) was fixed by one agent and then reviewed by independent adversarial reviewers until it came back clean. Integrations got a vendor-spec/auth lens and a canvas-vs-tool-call/backward-compat lens; batches got the canvas-vs-tool-call lens.
  • Regression tests call each mapper the way the Agent path does, params(modelArgs, { operation }) without an operation in params, and were confirmed to fail on the pre-fix code.
  • Ran the touched test files, biome, check:mapper-operation, check:explicit-any, check:tool-metadata, check:docs, check:docs-manifest, check:block-successors, check:deployment-config, check:canvas-sentences, check:test-patterns, check:utils, check:comment-hygiene, check:source-conventions and check:integration-catalog locally. CI runs type-check and the full suite.

🤖 Generated with Claude Code

@vercel

vercel Bot commented Nov 16, 2026 •

Copy link
Copy Markdown

The latest updates on your projects. Learn more about Vercel for GitHub.

1 Skipped Deployment
Project Deployment Actions Updated
docs Skipped Skipped Nov 16, 2026 3:10am UTC

Request Review

@cubic-dev-ai cubic-dev-ai Bot left a comment •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All reported issues were addressed across 239 files

Note: This PR contains a large number of files. cubic selects up to 200 of the highest-priority eligible files for this review, so some files may not have been reviewed.

Turn on auto-fix | Re-trigger cubic

Comment thread apps/sim/tools/gitlab/get_project.ts Resolved
Comment thread apps/sim/tools/linear/get_issue.ts Resolved
Comment thread apps/sim/tools/datadog/types.ts Resolved
Comment thread apps/sim/tools/sentry/projects_list.ts Resolved
@greptile-apps

greptile-apps Bot commented Nov 16, 2026 •

Copy link
Copy Markdown
Contributor

RetriggerView in GreptileConfidence Score: 5/5

[High impact] The changes since the previous review appear safe to merge.

Summary

This PR fixes request formats, field mappings, pagination, and outputs across eight integrations.

  • Agent tool calls now use the operation and field mappings saved in each block.
  • GitHub blocks send corrected requests and return more complete repository data.
  • GitLab blocks map more fields and build safer requests for project and user operations.
  • Linear blocks use supported operations and expose project status and notification results.
  • Slack tool calls accept more valid inputs and place limits on history reads.
  • incident.io action and follow-up lists use v3 and can return more pages of data.
  • PagerDuty incident and event calls handle saved fields and optional event details.
  • Sentry and Datadog blocks return useful paging data and accept more API-shaped inputs.

Reviews (3) · Last reviewed commit: "fix(integrations): keep latest-commit v2..." · Reviewed by Greptile

@waleedlatif1

Copy link
Copy Markdown
Collaborator Author

@greptile

@waleedlatif1

Copy link
Copy Markdown
Collaborator Author

@cubic-dev-ai review this PR

@cubic-dev-ai

cubic-dev-ai Bot commented Nov 16, 2026

Copy link
Copy Markdown
Contributor

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

@cubic-dev-ai cubic-dev-ai Bot left a comment •

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All reported issues were addressed across 240 files

Note: This PR contains a large number of files. cubic selects up to 200 of the highest-priority eligible files for this review, so some files may not have been reviewed.

Turn on auto-fix | Re-trigger cubic

Comment thread apps/sim/tools/linear/types.ts Resolved
@waleedlatif1

Copy link
Copy Markdown
Collaborator Author

@greptile

@waleedlatif1

Copy link
Copy Markdown
Collaborator Author

@cubic-dev-ai review this PR

@cubic-dev-ai

cubic-dev-ai Bot commented Nov 16, 2026

Copy link
Copy Markdown
Contributor

@cubic-dev-ai review this PR

@waleedlatif1 I have started the AI code review. It will take a few minutes to complete.

@cubic-dev-ai cubic-dev-ai Bot left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

No issues found across 241 files

Confidence score: 5/5

  • Automated review surfaced no issues in the provided summaries.
  • No files require special attention.

Note: This PR contains a large number of files. cubic selects up to 200 of the highest-priority eligible files for this review, so some files may not have been reviewed.

Turn on auto-fix | Re-trigger cubic

…l-call context and validate integrations against vendor specs
- keep the Linear project-status tool's display name so its docs anchor survives
- pass Slack's comma-separated ID lists through the workflow-block mapper
- reject '..' segments in GitHub content paths; only exact dot-segment GitLab ids
- reject non-array commits in Sentry release create
- mark nullable outputs (Linear dueDate/status, GitHub latest commit v2)
- align descriptions with the vendor contracts (incident.io, Slack, GitHub, Linear)
- drop a dead credential fallback the canonical-param check flags; fix test types
…fecycle timestamps nullable

Inline the v2 commit output so the docs generator still documents it, and shrink the unused-exports baseline now that COMMIT_DATA_OUTPUT_PROPERTIES is imported.
@waleedlatif1
waleedlatif1 merged commit 2ed6201 into staging Nov 16, 2026
84 checks passed

This branch was previously deployed

1 inactive deployment
Preview — bff8a9e1 Deployed Nov 16, 2026 by vercel[bot]
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.

1 participant