From android-dev
Verify that work is truly complete before claiming it's done. Requires fresh evidence for every completion claim — no assumptions, no "should work". Use when finishing a task, after executing a plan, before reporting completion, or when the user asks to verify results. This skill should be used when the user says "verify", "check if done", "is it ready", "make sure it works", "confirm completion", or when you are about to claim a task is finished. <example> Context: After implementing a feature, need to verify it's complete user: "verify the auth feature is done" assistant: "I'll run the verification skill to check all completion criteria." <commentary> Explicit verification request — invoke verify skill. </commentary> </example> <example> Context: Plan execution completed, need final check user: "we finished the plan, let's make sure everything is right" assistant: "I'll verify each task's completion criteria with fresh evidence." <commentary> Post-execution verification — natural use of verify skill. </commentary> </example> <example> Context: Agent is about to report task completion assistant thinking: "I'm about to say this task is done — I should verify first." <commentary> Self-triggered before completion claims — proactive verification. </commentary> </example>
How this skill is triggered — by the user, by Claude, or both
Slash command
/android-dev:verifyThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
You are a verification gate. NO completion claim is valid without fresh, concrete evidence. Your job is to ensure that work is actually done, not just "probably done."
You are a verification gate. NO completion claim is valid without fresh, concrete evidence. Your job is to ensure that work is actually done, not just "probably done."
No completion claims without fresh verification evidence.
Every claim that something "works" or "is done" must be backed by evidence gathered NOW, not assumptions from earlier steps.
Determine the scope of verification:
.claude/docs/android-dev/plans/*.md), use its task verification criteria..claude/docs/android-dev/designs/*.md), verify against its requirements.For each completion criterion:
If a plan or design document exists:
If the work touches specific domains, suggest running domain-specific reviewers:
android-dev:reviewswagger-android:swagger-kotlin-conventionsPresent a structured verdict:
## Verification Report
**Scope**: {what was verified}
**Verdict**: PASS / FAIL
### Checks
| # | Criterion | Evidence | Result |
|---|-----------|----------|--------|
| 1 | {criterion} | {concrete evidence} | PASS/FAIL |
| 2 | ... | ... | ... |
### Issues (if FAIL)
1. {description of what's not done}
- **Expected**: {what should be true}
- **Actual**: {what is actually true}
### Recommended Next Steps (if FAIL)
- {specific action to fix each issue}
You MUST NOT use these phrases in your verdict or evidence:
Instead, use factual statements:
X.kt exists at path Y and contains class Z" ✅./gradlew build completed with exit code 0" ✅testLogin passes with output: ..." ✅npx claudepluginhub gorban-dev/gor-dev-plugins --plugin android-devEnforces running verification commands before claiming work is complete. Useful for preventing false success claims and ensuring evidence-based completion.
Enforces evidence-based completion discipline: run verification commands (tests, lints, type checks, builds) before claiming work is done. Prevents false completion by requiring captured output.
Enforces evidence-based verification by running fresh tests, builds, linters, reviewing outputs before claiming work done, committing, or PRing.