verify

name	verify
description	AI self-verification of implemented work across surfaces (extension, CLI, MCP, TUI, workflow). Use after implementation to verify code works in its runtime environment.

Verify

AI self-verification of implemented work. Goes beyond unit tests to verify code actually works in its runtime environment.

Usage

• /verify - auto-detect from recent changes
• /verify cli|tui|extension|mcp|workflow - explicit mode

Prerequisites

Mode	Required
cli	Bash tool only
tui	tests/PPDS.Tui.E2eTests/tools/tui-verify.mjs (@microsoft/tui-test, node-pty)
extension	src/PPDS.Extension/tools/webview-cdp.mjs (@playwright/test, @vscode/test-electron)
mcp	npx @modelcontextprotocol/inspector

If a prerequisite is missing, tell the user what to install and stop.

Process

1. Detect component

From $ARGUMENTS or recent changes: src/PPDS.Cli/Commands/->cli, src/PPDS.Cli/Tui/->tui, src/PPDS.Extension/->extension, src/PPDS.Mcp/->mcp, .claude/+scripts/->workflow. No match: ask.

2. Unit tests first

• CLI/TUI: dotnet test PPDS.sln --filter "Category!=Integration" -v q
• Extension: npm run test --prefix src/PPDS.Extension
• MCP: dotnet test --filter "FullyQualifiedName~Mcp" -v q

Don't waste interactive cycles on broken code.

3. CLI mode

dotnet build src/PPDS.Cli/PPDS.Cli.csproj -f net10.0
.\src\PPDS.Cli\bin\Debug\net10.0\ppds.exe <command> <args>

Verify exit 0, format correct, edge cases (empty/invalid/no-auth).

4. TUI mode

See REFERENCE.md "TUI Mode - worked example" for the Phase A/B/C commands. Phase B is mandatory when changed files touch src/PPDS.Cli/Tui/.

5. Extension mode

See REFERENCE.md "Extension Mode - worked example" for Phase A/B/C commands. Phase B is mandatory when query/data/panel files change.

6. MCP mode

npx @modelcontextprotocol/inspector --cli --server "ppds-mcp-server"

Per tool: valid->success shape; edge case->error handling; matches schema.

7. Workflow mode

Check 1 - Python tests (the full automation suite — hook behavior, skill content, bash portability):

python -m pytest tests/ scripts/ --import-mode=importlib -q

Check 2 - Hook scripts: for each surviving .py in .claude/hooks/ (except _pathfix.py): echo '{}' | python .claude/hooks/{hook}.py. Exit 0 or 2, never 1.

Check 3 - settings.json: parses; every hook command exists on disk.

Check 4 - Skill frontmatter: every .claude/skills/*/SKILL.md parses; name and description non-empty.

Check 5 - Agent frontmatter: every .claude/agents/*.md parses; tools present; names in known set (Read, Edit, Write, Glob, Grep, Bash, Agent, WebSearch, WebFetch, NotebookEdit, and Bash(pattern:*)).

Check 6 - Skill file refs: every referenced path exists.

8. Report

See REFERENCE.md "Report template". Include actual values.

Verify is one step in the usual shipping sequence (/gates -> /verify -> /pr), not the last one. On PASS, proceed to /pr. On FAIL, run /debug to root-cause it (Rule 5: verify reports, it does not fix); after the fix lands, re-run /verify before continuing.

Rules

1. Unit tests first.
2. Screenshots for visual changes - look at them, don't just take them.
3. Report actual values, not just pass/fail.
4. Prerequisites are hard gates.
5. Don't fix during verify - report problems for /debug.