Help us improve
Share bugs, ideas, or general feedback.
From claude-commands
Enforces the Three Evidence Rule to verify automatic integration behavior—configuration, trigger, and log evidence required for claims like hooks, webhooks, CI, cron, events, and middleware.
npx claudepluginhub jleechanorg/claude-commands --plugin claude-commandsHow this skill is triggered — by the user, by Claude, or both
Slash command
/claude-commands:integration-verificationThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
**Purpose**: When claiming ANY integration works, you MUST provide the Three Evidence Rule to prove automatic behavior, not just manual testing.
Enforces evidence-based verification by running fresh tests, builds, linters, reviewing outputs before claiming work done, committing, or PRing.
Verifies claims about tests, builds, verification, or code quality using bash/git commands like status, diff, ls, cat for concrete evidence in dev workflows.
Enforces verification gates: run fresh commands before claiming work is complete, tests pass, or bugs are fixed. Prevents confirmation bias.
Share bugs, ideas, or general feedback.
Purpose: When claiming ANY integration works, you MUST provide the Three Evidence Rule to prove automatic behavior, not just manual testing.
For ANY integration claim, provide ALL THREE:
| Evidence Type | What It Proves | Example |
|---|---|---|
| Configuration Evidence | Feature is enabled | Config file entry, env var, feature flag |
| Trigger Evidence | Automatic activation | Hook registration, event listener, cron entry |
| Log Evidence | Actually executed | Timestamped logs from automatic (not manual) run |
### Integration Claim: <what you're claiming works>
#### 1. Configuration Evidence
```
File:
<code or config showing automatic trigger mechanism>
Mechanism:
<timestamped log output from automatic execution>
Timestamp: Triggered by:
## Common Integration Claims That Require This Protocol
- "The hook is working"
- "The webhook fires automatically"
- "CI runs this on push"
- "The cron job executes"
- "The event listener handles this"
- "The middleware intercepts requests"
## Examples
### Good Evidence (Complete)
```markdown
### Integration Claim: Pre-commit hook runs fake code detection
#### 1. Configuration Evidence
```yaml
# .pre-commit-config.yaml
repos:
- repo: local
hooks:
- id: fake-code-check
name: Fake Code Detection
entry: python scripts/fake_detector.py
language: python
types: [python]
File: .pre-commit-config.yaml (line 12-18)
$ git config --get core.hooksPath
.git/hooks
$ ls -la .git/hooks/pre-commit
-rwxr-xr-x 1 user user 478 Jan 15 10:00 .git/hooks/pre-commit
Mechanism: Git pre-commit hook installed via pre-commit framework
2025-01-15T10:05:23 [pre-commit] Running fake-code-check...
2025-01-15T10:05:24 [fake_detector] Scanning 3 files...
2025-01-15T10:05:25 [fake_detector] No fake code detected
2025-01-15T10:05:25 [pre-commit] Passed fake-code-check
Timestamp: 2025-01-15T10:05:23
Triggered by: git commit -m "Add feature" (automatic, not manual script run)
### Bad Evidence (Incomplete)
```markdown
### Integration Claim: Pre-commit hook runs fake code detection
"I ran the script and it worked." <-- NO CONFIG EVIDENCE
"The config file has the hook." <-- NO TRIGGER EVIDENCE
"Here's manual test output." <-- NO AUTOMATIC LOG EVIDENCE
| Evidence Type | Valid | Invalid |
|---|---|---|
| Script run manually | No | "python script.py" output |
| Script triggered by hook | Yes | Log shows hook invocation |
| Curl request sent manually | No | Manual API testing |
| Webhook received from external | Yes | Log shows webhook payload |
When verifying work done by another agent or automated process:
ls -la <expected files>git diff --stat to see actual changesgit status for uncommitted workNever trust "it's done" without verification evidence.
Before claiming integration works:
If you cannot provide all three, the integration is NOT verified.