From memstack
Verifies completed work before committing: identifies task from context/git diff, runs project build/test/lint commands, manually checks requirements, scans for debug code and common issues.
npx claudepluginhub cwinvestments/memstack --plugin memstackThis skill uses the workspace's default tool permissions.
*Review completed work against requirements before committing.*
Enforces evidence-based verification by running fresh tests, builds, linters, reviewing outputs before claiming work done, committing, or PRing.
Enforces running fresh verification commands like tests, builds, or linters before claiming work is complete, fixed, passing, committing, or creating PRs.
Enforces five-step verification (identify/run/read/verify/claim) of commands like npm test/build/lint before claiming code complete, fixed, or passing. Builds trust via evidence.
Share bugs, ideas, or general feedback.
Review completed work against requirements before committing.
When this skill activates, output:
✅ Verify — Checking work against requirements...
Then execute the protocol below.
| Context | Status | Priority |
|---|---|---|
| User says "verify", "check this work", "does it pass" | ACTIVE — run verification | P1 |
| User says "is this done", "ready to commit" | ACTIVE — run verification | P1 |
| User is mid-task, still actively coding | DORMANT — let them finish first | — |
| User asks to commit or push | DORMANT — Seal hook handles pre-push checks | — |
| User asks to review someone else's code | DORMANT — not a code review tool | — |
If you're thinking any of these, STOP — you're about to skip the protocol:
| You're thinking... | Reality |
|---|---|
| "I already tested this manually" | Ad-hoc testing leaves no record. Run the automated checks. |
| "This change is too small to verify" | Small changes cause regressions. Run the full protocol. |
| "The build passed, so it's fine" | Build passing ≠ requirements met. Always do the manual requirement check (Step 3). |
| "I'll just commit and fix issues later" | Issues found after commit are 10x harder to fix. Verify first. |
| "There are no tests for this project" | Skip automated tests, but ALWAYS do the manual requirement check and common issues scan. |
| "The user seems in a hurry" | Shipping broken code wastes more time than verification takes. |
Determine what was being built by checking:
Summarize: "Task: [what was being built/changed]"
Run whatever applies to the current project:
Build check:
# Detect and run the project's build command
npm run build # Node.js
make build # Makefile
python -m py_compile <file> # Python
Test check:
# Detect and run the project's test command
npm test # Node.js
pytest # Python
make test # Makefile
Lint check (if available):
npm run lint # Node.js
If no build/test tooling exists, skip and note "No automated checks configured."
Compare the completed work against the original requirements:
console.log, print(), debugger, TODO markers left behind## Verification Report
*{YYYY-MM-DD}*
### Task: {what was being built}
### Automated Checks
- [x] Build: passes / [ ] fails — {error}
- [x] Tests: passes / [ ] fails — {error}
- [x] Lint: passes / [ ] N/A
### Requirements
- [x] {Requirement 1} — implemented in {file}
- [x] {Requirement 2} — implemented in {file}
- [ ] {Requirement 3} — MISSING: {what's needed}
### Issues Found
1. {Issue} → {Suggested fix}
### Verdict: PASS / NEEDS FIX
User: "verify this work"
✅ Verify — Checking work against requirements...
## Verification Report
Task: Add user registration API endpoint
### Automated Checks
- [x] Build: passes
- [x] Tests: 14/14 passing
- [x] Lint: clean
### Requirements
- [x] POST /api/register endpoint — server/routes/auth.ts
- [x] Email validation — uses zod schema
- [x] Password hashing — bcrypt with salt rounds=12
- [ ] Rate limiting — MISSING: no rate limiter on registration endpoint
### Issues Found
1. No rate limiting on /api/register → Add express-rate-limit middleware
### Verdict: NEEDS FIX (1 issue)