Know why the build misfired.
An open-source GitHub Action that, when a build goes red, runs an agentic AI investigation of the failure and posts an advisory PR comment — explaining the root cause and classifying the failure as real vs. likely flaky. Advisory only: it never blocks, merges, or changes code.
Bring your own key. Misfire runs on Claude or OpenAI (pick with
--provider; the OpenAI path also drives OpenAI-compatible endpoints like GitHub
Models / Copilot). It never provides an API key — you set your own, and every
analysis is billed to your account. The local --no-ai mode needs no key.
How it works
On a red build, Misfire runs a tool-using agent over four shared, mockable tools:
get_failing_logs— the failing CI job logsget_pr_diff— the change under testget_test_history— past pass/fail rate (powers the real-vs-flaky verdict)post_pr_comment— the single write action
Each provider is an adapter behind one seam (misfire/agents/) — Claude and
OpenAI today, more easily added — so the shared tools and prompt don't change
when you switch provider with --provider.
Usage
Add a workflow that runs after your CI workflow and comments on red builds.
Using the decoupled workflow_run trigger means Misfire gets a write-scoped
token (so it works on fork PRs) and can never block or slow your CI job:
# .github/workflows/misfire.yml
name: Misfire
on:
workflow_run:
workflows: ["CI"] # ← name of the workflow to watch
types: [completed]
permissions:
contents: read
pull-requests: write
actions: read
jobs:
explain:
if: github.event.workflow_run.conclusion == 'failure'
runs-on: ubuntu-latest
steps:
- uses: ericmhayes/misfire-ci@v0
with:
anthropic-api-key: ${{ secrets.ANTHROPIC_API_KEY }}
github-token: ${{ secrets.GITHUB_TOKEN }}
run-id: ${{ github.event.workflow_run.id }}
pr-number is optional: when omitted Misfire resolves the PR from the failed
run's head commit via the GitHub API. This works for PRs from forks too (where
workflow_run.pull_requests is empty), and it simply no-ops when the run has no
associated PR (e.g. a push straight to the default branch).
Add your ANTHROPIC_API_KEY as a repository secret (bring-your-own-key). The
built-in GITHUB_TOKEN supplies the pull-request-write / actions-read scopes.
Local / offline use
Analyze logs straight from a file or stdin — no GitHub, nothing posted. The result is printed to stdout.
# Deterministic, no model call, no key (the offline fallback):
misfire analyze --logs ci.log --no-ai
cat ci.log | misfire analyze --logs - --no-ai
# Full agentic diagnosis (needs ANTHROPIC_API_KEY), optionally with the diff:
misfire analyze --logs ci.log --diff change.diff
--no-ai gives a deterministic rule-based summary (failure category + the key
log excerpt); the AI path runs the same agent and prints the advisory instead of
posting it. Secrets in the logs/diff are scrubbed before anything is sent to the
model.
Install (development)
git clone https://github.com/ericmhayes/misfire-ci.git
cd misfire-ci
pip install -e ".[dev]"
License
MIT — bring-your-own-key (you supply your own Anthropic API key).
Metadata
Release files for misfire-ci 0.1.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| misfire_ci-0.1.0.tar.gz | 44.4 kB | With provenance | Details for misfire_ci-0.1.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| misfire_ci-0.1.0-py3-none-any.whl | Python 3 | none | any | Details for misfire_ci-0.1.0-py3-none-any.whl |
Total release size: 80.6 kB
Release files / misfire_ci-0.1.0.tar.gz
| Download URL | misfire_ci-0.1.0.tar.gz |
|---|---|
| Size | 44.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
29cd7857a208eaa17b822f373c24df97d987cf810244979af73ef619fb926ba0
|
|
BLAKE2b-256 checksum How to use checksums |
c443d64c9ac6d03d3d3d122d010439817c706e137edd8a9177946a9bb891fd2c
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 21, 2026.
Transparency logRelease files / misfire_ci-0.1.0-py3-none-any.whl
| Download URL | misfire_ci-0.1.0-py3-none-any.whl |
|---|---|
| Size | 36.2 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
0d7f0658ce744b8c77b8dab9f8e88ad651e899c1e35232cb8a2ccf8b379891da
|
|
BLAKE2b-256 checksum How to use checksums |
420f70ac53950b1a0b4f8786d3ffd195d1f0330a560cf22884b1e62e62f8a481
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/7.0.0 CPython/3.13.16
|
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 21, 2026.
Transparency logSecurity for misfire-ci 0.1.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
Full provenance
Both files in this release have been uploaded with provenance metadata.
Attestation for misfire_ci-0.1.0.tar.gz
PyPI Publish Attestation
PyPI verified that this artifact, at this checksum, originated from the publisher listed below.
Signed by GitHub Actions, verified by PyPI on Nov 21, 2026.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.