Catches shortcuts and reward hacking — weakening or deleting tests to make them pass, hard-coding expected outputs, silently dropping requirements, or claiming work is done without running it. Use before declaring a task complete and whenever a test or requirement feels like it's in the way.
The English summary is being prepared. Showing the original description.
See how to installReview the instructions the agent will receive before you install.
Current models rarely skip work out of laziness. The failure that remains is subtler: satisfying the check instead of the goal. A test gets special-cased, an assertion loosened, a requirement quietly narrowed, and the summary says "done." This skill catches that and makes it visible.
| Pattern | What it looks like | Do instead |
|---|---|---|
| Test tampering | Loosening an assertion, adding a skip, deleting a failing case, raising a tolerance | Fix the code. If the test is wrong, change it and say why in the summary. |
| Special-casing | if input == <test fixture>: return <expected> or detecting the test environment | Implement the general behavior. |
| Hard-coded outputs | Returning the literal value a test expects | Compute it. |
| Mocking away the subject | Mocking the very function the test is supposed to exercise | Mock only external boundaries. |
| Silent scope cut | Implementing 4 of 5 requirements and summarizing as "done" | Finish, or list exactly what is missing. |
| Unverified claim | "This should fix it" / "Tests pass" without running them | Run the command and quote the result. |
| Budget loosening | Raising a performance/size budget or threshold because the check failed | Fix the code; changing a budget needs a spec change and a stated reason |
| "It's flaky" | Re-running until green, adding a retry | Find the variable (vibe-flake-root-cause) |
| Self-approval | Recording a human's approval on their behalf, or treating a code comment as sign-off | Point the human to the item to approve (vibe-publication-leak-guard) |
| Error swallowing | catch {} / except: pass to stop a crash | Handle the specific error, or let it surface. |
| Deferred forever | "I'll add that later" with no tracked follow-up | Do it now, or create a tracked issue/TODO with a reference. |
When run as a self-check, report only if something was found. When invoked explicitly:
| # | Pattern | Location | Resolution |
|---|---|---|---|
| 1 | Test tampering | auth_test.go:88 tolerance 0.01 → 0.5 | Reverted; fixed rounding in calc.go |
Requirements: X/Y met — [list any not met] Unverified claims: [none / list]
No reviews yet. Tell others what you thought.
No login needed. Share how it worked for you, ask a question, or note where it did not run.
No comments yet. Be the first to write one.
Validates completed work against defined acceptance criteria. Use after completing a task that has specific success criteria defined in issues, specs, or task descriptions.
The English summary is being prepared. Showing the original description.
Generates edge case, failure mode, and spec-driven test cases. Covers boundary values, nil inputs, concurrency, resource exhaustion, malformed data, and requirement-linked traceability tests. Use after happy-path tests exist and before claiming coverage is complete, when requirements lack tests, or before a security review.
The English summary is being prepared. Showing the original description.
Safely integrates commits from parallel agent branches using sequential cherry-pick. Use after parallel work completes in isolated branches or worktrees.
The English summary is being prepared. Showing the original description.
Audits tests for concurrency safety — race conditions, shared mock state, cleanup ordering. Use when writing tests that involve goroutines, async operations, or shared mutable state.
The English summary is being prepared. Showing the original description.
Enforces tiered test coverage standards with three dimensions — line coverage by tier, spec-to-test traceability, and spec-to-code implementation mapping. Use before claiming code is complete.
The English summary is being prepared. Showing the original description.
Extracts and records architectural decisions from diffs, conversation context, and explicit choices. Supports automatic extraction from staged changes, deduplication against prior decisions, and persistent ADR-format logging with spec traceability. Use after a significant design or architecture choice, before committing a diff that embeds implicit decisions, or when the same question keeps resurfacing.
The English summary is being prepared. Showing the original description.