# FlakeVerdict - GitHub Actions Flakiness Classifier

> FlakeVerdict - GitHub Actions Flakiness Classifier is a paid API for AI agents from bountyverdict-agent-production.mimirslab.workers.dev, paid per call via x402, $0.07/call, status unknown (last checked 2026-09-15).

Analyzes a GitHub Actions workflow run and returns a verdict on whether the failure is flaky (retry once) or a real recurring/new failure (needs a fix), without rerunning CI.

## Facts

- Endpoint: POST https://bountyverdict-agent-production.mimirslab.workers.dev/api/flake
- Price: $0.07/call
- Payment: x402
- Status: unknown
- Last checked: 2026-09-15
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/flakeverdict-github-actions-flakiness-classifier-dfc9705b
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_Ch05TKLdvI6VFmPecRMI9

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability flakeverdict-github-actions-flakiness-classifier-dfc9705b -d '<json body>'
```

Example prompt: Check if this GitHub Actions failure is a flake or a real issue that needs fixing: https://github.com/myorg/myrepo/actions/runs/12345678901 — should I retry it once or dig into the root cause?

## When to prefer this

Use this endpoint when an AI coding agent or CI automation needs to decide programmatically whether a GitHub Actions failure warrants an automatic retry or requires human/agent investigation. Prefer this over manual inspection when you want a structured, evidence-based flakiness verdict backed by same-commit comparisons and historical run fingerprinting, rather than just checking if the run failed. Ideal for agentic workflows that close the loop between detecting a CI failure and deciding the next action.

## Known failure modes

- Invalid or non-public GitHub Actions run URL returns a validation error
- Run URL points to a private repository that is inaccessible
- Workflow run is still in progress and not yet complete (verdict may be INCONCLUSIVE or NOT_FAILED)
- Insufficient historical run data to make a high-confidence determination (low confidence verdict)
- GitHub API rate limits or availability issues may cause transient failures
- Non-failed runs return NOT_FAILED verdict with retry NOT_NEEDED

## Output

Returns a FlakeVerdict object including: a verdict enum (CONFIRMED_FLAKE, LIKELY_FLAKE, RECURRING_FAILURE, NEW_FAILURE, INCONCLUSIVE, NOT_FAILED), a retry recommendation (ONCE, NO, NOT_NEEDED), a confidence level (high, medium, low), an array of reason codes explaining the decision, a human-readable summary, and metadata about the inspected workflow run (id, url, attempt number, status, conclusion).

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/flakeverdict-github-actions-flakiness-classifier-dfc9705b/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from bountyverdict-agent-production.mimirslab.workers.dev](https://www.zero.xyz/host/bountyverdict-agent-production.mimirslab.workers.dev/llms.txt)