Publisher: Jesse Vincent / Prime Radiant · Editorial profile by AI Tools Ratings · Review method
What task is this skill designed for?
Review a synthetic change whose unit tests pass but whose build fails, then produce an accurate completion report.
What input and output should you expect?
Input: The intended deliverable, current source tree, test and build commands, and any acceptance conditions that cannot be checked automatically.
Output: A concise claim-to-evidence report showing what passed, what failed and which requirements remain unresolved. This is a reporting workflow, not a substitute for missing tests.
What does it need to run?
A coding agent able to execute the project verification commands and inspect full output and exit codes. Verification may require browsers, services or credentials that are not present.
Publisher README documents Claude Code, Codex and Cursor installation paths, among other harnesses. Tool availability and behavior depend on the host; this is publisher-declared support, not a cross-platform test result.
How should you approach installation?
- Review the pinned folder and its referenced files before enabling the workflow. The folder belongs to the broader Superpowers plugin, not an independently verified standalone installer.
- Follow the publisher README for your chosen coding-agent platform. The publisher documents several harnesses; use the instructions for your installed version and review plugin hooks and permissions.
- Try the suggested task in a disposable repository with synthetic inputs. Check the acceptance list before applying the workflow to shared work.
Installation guide: discovery, dependencies and runtime checks
What files, accounts or data can it access?
Runs commands and reads code, results and logs. Tests can write files or invoke services; the Skill itself does not make an unsafe test command safe.
What does the skill cost?
No separate skill purchase is specified in the MIT-licensed source. Coding-agent subscriptions, model tokens, test compute and any external services have their own charges. Review or repeated execution can increase model usage.
What are its limits and alternatives?
Fresh evidence only covers the tested tree and conditions. A green unit suite does not establish deployment health, accessibility or real user outcomes. Expensive full-suite checks should be planned within the project workflow.
Use a CI gate for repeatable automated enforcement. Use this Skill to make the agent handoff accurately reflect both automated results and checks that require human review.
What could you ask in a small trial?
Try this with a small sample:
Check this change before calling it complete. Map each acceptance condition to evidence, run the available checks and report the failing build even if the unit suite passes.
How would you verify the result?
- The final statement identifies the failed build and does not label the overall change complete.
- Each passing claim names the command and observed result.
- Unavailable manual or environment-dependent checks remain explicit rather than inferred from a green test.
Use a baseline and record acceptance evidence before treating a successful task as a reliable workflow.
Which source and license were checked?
Read the reviewed SKILL.md at its pinned commit · Complete reviewed folder
Repository and current publisher guidance
Checked:
MIT — repository LICENSE at the reviewed commit. Dependencies and external services retain their own terms. License evidence
Version details and review scope
Source: obra/superpowers/skills/verification-before-completion
Reviewed commit: 2ce8aa15d05bd57213e6f746f953aafb7ded7a8b
SKILL.md SHA-256: 2befe7fc55bcadaa3d97dd9e8efeb633d2561c0ebe74c5a8b17c4d9e7e4520b3
Source-document review. This profile covers the source workflow, stated dependencies and license evidence. See our editorial methodology for how we distinguish source reviews and hands-on evaluations.