Agent template
CI Failure Investigator
On a failed GitHub Actions run, fetches the workflow logs, classifies the failure (flake / real / infra), and posts a diagnosis comment on the PR so reviewers know whether to retry or fix.
template · ci-failure-investigator
- Category
- Operations
- Setup
- Advanced setup
- Connects to
- GitHub
- Tools
- 6
- Model
- gpt-4.1-mini
What it uses
The tools this agent may call.
- Current time
- GitHub: get a workflow run
- GitHub: list pull requests
- GitHub: get a pull request
- GitHub: create an issue comment
- Read a web page
Any of them can require a person's approval before it runs. You choose which, after you install it.
Getting it running
Connect its apps and fill in the parts of the prompt that describe your team.
Install it
It becomes a regular agent in your workspace, yours to change.Connect GitHub
Sign in to each app once; every agent can then use it.Choose the gates
Pick which of its tools wait for a person's approval.Run it
On demand or on a schedule, and read the first run end to end.
Install it in your workspace
It runs on your model keys, with zero markup on tokens.
More in Operations
Other agents for the same work.
- OperationsPR Reviewer BotWatches new GitHub pull requests, drafts a review summary with risk flags, and posts it to a Slack channel, pinging humans on large or high-risk diffs.Advanced setupGitHubSlack
- OperationsIssue TriageReads new GitHub or Linear issues, applies labels (type, area, priority), checks for duplicates against existing issues, and comments with a triage summary.Advanced setupGitHub
- OperationsStandup DigestPulls each engineer's in-progress Linear issues and posts a per-team daily standup digest to Slack at 9am, who's on what, what's blocked, what shipped yesterday.Advanced setupLinearSlack
Integrations
18 apps, your models, your APIs.
Connect GitHub, Gmail, Slack and more once, describe any HTTP API as a tool, and choose the model per agent.
Install it, read its first run.
Every template is a regular agent once installed: change its prompt, its tools and its approvals, and see every run.