Skip to content

About

A skill for reviewing software plans and release candidates, with findings tied to evidence.

Topics

Resources

Contributing

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

Engineering Assurance

Engineering Assurance is an agent skill for planning a software project before most of it is built, and for reviewing a specific commit before you release it. It's one coordinating skill plus procedures it only loads when a review needs them, and some Python helpers that check the report it writes.

It's built for a review someone is watching, not an unattended pipeline step. For a bigger review it can split the work across code, security, runtime and delivery procedures. It works within your product decisions and your project's standards, and it keeps what it found separate from what it couldn't check. Whether to ship is still your call.

Start here

Load the whole skills/engineering-assurance/ directory the way your harness loads skills. SKILL.md on its own isn't enough, because it reads the files next to it. The usage guide covers loading the skill and choosing the work, including reports and optional reproducible packets. Pin the commit of this repo you reviewed, and record it separately from the commit you're reviewing.

Example request:

Use Engineering Assurance in prerelease audit mode on this commit. Review the complete candidate within the agreed scope, return findings and coverage gaps, and keep fixes as proposals.

There are two phases, prebuild and prerelease, and four modes: audit, remediation-planning, authorized-remediation and reverification. On a new project or a prototype, use prebuild with your accepted product design. If you've already agreed which fixes to make, use authorized-remediation, then reverification to check the result.

The helpers and tests need Python 3.11 or later and Git, and nothing from pip.

python3 -m unittest discover -s tests
python3 scripts/check_package.py
python3 skills/engineering-assurance/scripts/validate_report.py --help

The record contract describes what a report must contain. examples/report.json is a made-up report with one finding and one coverage gap left in on purpose. The optional GitLab observation job checks supplied reports in a consuming project.

None of these checks tell you whether the agent found every defect, or whether your product is ready to ship.

See contributing for the change workflow and the changelog for what's in the first release. Credits links the projects that informed the procedures.

License

Copyright 2026 Maxfield Allison. Engineering Assurance is licensed under the Apache License 2.0.

About

A skill for reviewing software plans and release candidates, with findings tied to evidence.

Topics

Resources

Contributing

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages