Code Verification Pipeline

Code Verification Pipeline

Overview

Orchestrates a deep verification pass that checks whether code in scope is correct, clean, well-tested, and functional. This pipeline reports findings — it does NOT auto-fix anything. The user decides what to fix based on the report.

When to Use

• User asks to verify, run tests, lint, type-check, or gate quality.
• Triggers: "verify", "run tests", "check code quality", "lint the code", "validate".
• You need a read-only pass/fail snapshot before shipping or reviewing.

When NOT to use

• Failing tests that need fixing — use polish or debug.
• New feature work that includes tests — use build.
• Bug hunt with auto-fix — use audit.

Input

• $ARGUMENTS — Optional scope and focus. Any combination of:
  • Path/glob: "src/auth/", "tests/test_payments.py" — which files to verify
  • Free-text focus: "check examples work", "verify the API layer" — what to focus on
  • Combined: "src/payments/ check all edge cases are tested"

If empty, verify the entire project (roam mode with standard exclusions).

Scope Parameter

See skills/shared/scope-parameter.md for the canonical scope modes (branch, global, working) and their git commands. Document any skill-specific overrides or restrictions below this line.

• Default (no scope:): path/glob detection from argument, or roam entire project if empty.
• scope: determines which source files are under verification. Tests always run fully (not filtered) to catch regressions. Coverage and quality analysis focus on code within scope.

Constants

• TASK_DIR: .mz/task/ — working artifacts under .mz/task/<task_name>/
• MAX_RESEARCHERS: 3 — for failure diagnosis and coverage/quality review

Core Process

Phase Overview

#	Phase	Reference	Loop?
0	Setup	inline below	—
1	Scope Resolution	phases/setup.md	—
2	Tooling Detection	phases/setup.md	—
3	Execution	phases/checks.md	—
4	Coverage & Quality Analysis	phases/checks.md	—
5	Failure Diagnosis	phases/checks.md	—
6	Report	phases/checks.md	—

Phase 0: Setup

1. Parse argument — split (after removing scope:) into path-like tokens (globs, dirs, files) and focus tokens (free text).
2. Task name — <YYYY_MM_DD>_verify_<slug> where <YYYY_MM_DD> is today's date (underscores) and slug is snake_case (max 20 chars); on same-day collision append _v2, _v3.
3. Task dir & state — create .mz/task/<task_name>/, write state.md with Status, Phase, Started.
4. Task tracking — TaskCreate per pipeline phase.

Phase 1–6

• Phase 1 — Scope Resolution: resolve argument into source, test, and example files. See phases/setup.md → Phase 1. Update state to scope_resolved.
• Phase 2 — Tooling Detection: detect test frameworks, linters, formatters, type checkers, example runners. See phases/setup.md → Phase 2. Update state to tooling_detected.
• Phase 3 — Execution: run all detected tools, capture results. See phases/checks.md → Phase 3. Update state to checks_executed.
• Phase 4 — Coverage & Quality Analysis: dispatch coverage and quality reviewer agents for code in scope. See phases/checks.md → Phase 4. Update state to analysis_complete.
• Phase 5 — Failure Diagnosis: if any Phase 3 check failed, dispatch researchers; skip if green. See phases/checks.md → Phase 5. Update state to diagnosis_complete.
• Phase 6 — Report: compile results, write to .mz/reports/<YYYY_MM_DD>_test_<scope_name>.md (append _v2, _v3 if exists). See phases/checks.md → Phase 6. Update state to completed.

Techniques

Techniques: delegated to phase files — see Phase Overview table above.

Common Rationalizations

Rationalization	Rebuttal
"tests passed last time, skip rerun"	"environment drift silently breaks suites between runs"
"type-check is slow, skip"	"the bug you didn't type-check is the one that crashes in staging"
"coverage is a vanity metric"	"coverage < 70% means you're shipping blind in the uncovered 30%"

Red Flags

• Type-check was skipped because it was "slow".
• Tests were declared passing without running them fresh in this session.
• Coverage delta was not measured against the scope.

Verification

Output the final report block: per-check pass/fail status, coverage percentages, top quality findings, and the written report path. No silent skips.

Error Handling

• No test framework detected: ask the user for a test command. Do not skip.
• No linter detected: note it in the report, skip lint checks.
• No formatter detected: note it in the report, skip format checks.
• No type checker configured: note it in the report, skip type checks. Do NOT suggest adding one.
• Test command times out: report timeout with partial output, flag in report.
• Example script crashes: capture the error output, include in report, do not retry.
• Ambiguous scope: ask the user to clarify.
• Empty scope: report and exit.

State Management

After each phase, update .mz/task/<task_name>/state.md with:

• Current phase
• Per-check results (pass/fail/skip)
• Any issues encountered