audit-python-code

{!% require_skill 'python:python-code-standards' %!}

<objective>

A verdict on Python implementation code — APPROVED, or REJECTED with each finding naming the design flaw, the violated rule, and the evidence.

</objective>

<constraints>

• NEVER modify files, generate fixes, write replacement code, commit changes, or change project state — this audit produces a verdict only.
• NEVER run deterministic validation, lint, type-check, test, or eval commands — this audit reads and judges; it never runs deterministic verification.
• NEVER evaluate test evidence quality — /audit-python-test is the separate concern that judges it.
• ALWAYS keep findings to artifact, violated rule, evidence, and why the cited code violates the rule.
• NEVER include corrective code samples, implementation patches, prescribed refactors, or required-change summaries in the verdict.
• APPROVED means every concern row passes or is explicitly not applicable. REJECTED means at least one concern row fails. APPROVED output contains no notes, warnings, or suggestions sections.

</constraints>

<repo_local_overlay> Standards are pre-loaded above. Check for spx/local/python.md at the repository root. Read it if it exists and apply it as repo-local routing to the product's governing specs and decisions. A local overlay supplements skill behavior; it does not declare product truth. </repo_local_overlay>

<essential_principles>

Comprehension is the whole job.

This audit reads and judges Python implementation code; it runs no deterministic verification. Do NOT run or re-check the project's linters, type-checker, or tests. Spend the whole audit on comprehension.

Comprehension is the core value.

Automated tools catch syntax errors, type mismatches, and lint violations. Claude catches: functions that do more than their name says, dead parameters required by no Protocol, IO tangled with logic, and designs that will break under change. The predict/verify protocol (Phase 1) is how these surface.

Test evidence is out of scope.

/audit-python-test evaluates whether tests provide behavior-coupled evidence using the 4-property model (coupling, falsifiability, alignment, coverage). This skill judges implementation design, not test evidence, and never runs the test suite. Do not duplicate that work.

Binary verdict, no caveats.

The verdict is the only output. Findings prove violations; they do not prescribe fixes.

</essential_principles>

<audit_workflow>

Execute phases IN ORDER. Do not skip. This audit runs no deterministic verification — no linter, type-checker, or test run.

Phase 0: Scope and Product Config

1. Determine target files/directories
2. Check pyproject.toml and {{! file('root_guide') !}}/README.md for tool and project configuration that informs comprehension (ruff, mypy, pytest settings; naming conventions) — read for context, never to run a gate. The linters already handled type annotations, magic numbers, bare excepts, unused imports, commented-out code, modern syntax, and security rules; comprehension covers what they cannot — deep relative imports, sys.path manipulation, unqualified Any, and # type: ignore without justification.

Phase 1: Code Comprehension

Read every file. Understand it. Question it. Do NOT skim, sample, or check boxes.

1.1 Per-Function Protocol

For each function/method:

1. Read name and signature only -- name, parameters, return type
2. Predict what it does in one sentence
3. Read the body -- validate the prediction
4. Investigate surprises:

Surprise	What it suggests
Parameter never used in body	Dead parameter -- required by Protocol, or remove
Does more than name says	SRP violation or misleading name
Does less than name says	Name overpromises or logic is incomplete
Variable assigned but never read	Dead code or unfinished logic
Code path that can never execute	Dead branch given calling context
Return value contradicts the type hint	Logic error or wrong return type

Prediction matched? Move on. Surprise? Document it with file:line.

1.2 Design Evaluation

For the codebase as a whole:

• IO vs logic separation -- Can core logic be tested without IO? Tangled computation and side effects need factoring.
• Dependency injection -- External dependencies injected via parameters or Protocol, or imported as globals?
• Single responsibility -- Each module/class does one thing? Each function does one thing?
• Error quality -- Errors include what failed and with what input?
• Domain exceptions -- Custom exceptions for domain errors, or everything generic ValueError/RuntimeError?

1.3 Import Evaluation

Evaluate import structure using the same vocabulary as /audit-python-test:

Import pattern	Classification
import pytest	Framework -- not reviewed
from hypothesis import given	Framework -- not reviewed
import json	Stdlib -- not reviewed
from typing import TYPE_CHECKING	Type-only -- erased at runtime
from product.config import parse_config	Codebase (production) -- review
from ..config import parse_config	Codebase (relative) -- review
from product_testing.harnesses import ConfigTestHarness	Codebase (test infra) -- review

Import depth rules:

Depth	Example	Verdict
Package import	from product.config ...	OK -- preferred
1 level	from ..config ...	Review -- truly module-internal?
2+ levels	from ....helpers ...	REJECT -- use package import
sys.path manip.	sys.path.insert(0, ...)	REJECT -- always

For stable locations (product_testing.harnesses.*, product_testing.generators.*, and inert fixture path providers), package imports are mandatory.

See ${CLAUDE_SKILL_DIR}/references/false-positive-handling.md for application context when evaluating security and linter suppression comments.

Phase 2: ADR/PDR Compliance

Find applicable ADRs/PDRs in the spec hierarchy (*.adr.md, *.pdr.md). Verify each constraint is followed. Undocumented deviations make this concern FAIL. If the product has no spec hierarchy, this concern is NOT_APPLICABLE.

Decision Record Constraint	Violation Example	Verdict
"Use dependency injection" (ADR)	Direct imports of external services	FAIL
"l1 tests for logic" (ADR)	l1 tests hitting network	FAIL
"No ORM" (ADR)	SQLAlchemy models introduced	FAIL
"Lifecycle is Draft→Published" (PDR)	Added hidden Archived state	FAIL

</audit_workflow>

<failure_modes>

These are real failures from past audits. Study them to avoid repeating them.

Approved code that passed ruff+mypy but had a design flaw. Claude trusted the green linters and skimmed comprehension. The code had a function named validate_config that also wrote the config file -- SRP violation hidden behind a reasonable name. The predict/verify protocol would have caught it: "Given the name, I predict this validates. But the body also calls Path.write_text(). Surprise."

Rejected code for a false positive. Claude flagged a parameter as "dead code" because it wasn't used in the function body. The parameter was required by a CommandHandler Protocol contract -- other implementations used it. Before flagging dead parameters, check if the function implements a Protocol.

Tried to evaluate test evidence instead of delegating. Claude found lambda cmd: (0, "", "") in tests and spent time analyzing whether it severed coupling. That's /audit-python-test' job, and this audit never runs the test suite. Claude should have moved straight to comprehending the implementation code.

Distracted by style while missing a logic bug. Claude spent review time on naming conventions, import ordering, and docstring completeness. Meanwhile, a branch condition was inverted -- if is_valid should have been if not is_valid. Comprehension (understanding what the code does) must come before style. Style is the linter's job.

Accepted code with tangled IO. A process_orders function both computed order totals AND sent confirmation emails. Tests passed and types were correct. But the function was untestable without an email server -- IO and logic were tangled. The design evaluation (1.2) would have caught it: "Can core logic be tested without IO? No."

</failure_modes>

<verdict_format>

Emit a structured verdict. The skill's entire output is the verdict payload.

The skill's overall is APPROVED iff every concern row is PASS or NOT_APPLICABLE; it is REJECTED if any concern is FAIL. An unavailable required inspection is FAIL, never NOT_APPLICABLE. Findings use severity blocking or debt.

{
  "schema_version": 1,
  "skill": "audit-python-code",
  "target": "<scope-target>",
  "overall": "APPROVED | REJECTED",
  "rows": [
    { "name": "function-comprehension", "status": "PASS | FAIL | NOT_APPLICABLE", "explanation": "<required when NOT_APPLICABLE>", "findings": [] },
    { "name": "design-coherence", "status": "PASS | FAIL | NOT_APPLICABLE", "explanation": "<required when NOT_APPLICABLE>", "findings": [] },
    { "name": "import-structure", "status": "PASS | FAIL | NOT_APPLICABLE", "explanation": "<required when NOT_APPLICABLE>", "findings": [] },
    { "name": "adr-pdr-compliance", "status": "PASS | FAIL | NOT_APPLICABLE", "explanation": "<required when NOT_APPLICABLE>", "findings": [] }
  ],
  "metadata": { "branch": "<branch>" }
}

Each finding carries file, line, rule (the concern name from the verdict table or a specific violation name), severity: "blocking | debt", message, observed, and expected. The message names the violated rule and consequence only; it contains no corrective code sample, implementation patch, prescribed refactor, or required-change summary.

</verdict_format>

<what_to_avoid>

• Do NOT run or re-check the project's linters, type-checker, or tests — this audit runs no deterministic verification
• Do NOT evaluate test evidence quality; /audit-python-test is the separate concern that judges it
• Do NOT commit or modify code (this skill is read-only)
• Do NOT generate fixes, replacement code, refactors, or required-change summaries
• Do NOT approve with caveats (binary verdict only)
• Do NOT reject for code style when comprehension found no design flaws

</what_to_avoid>

<example_review> Read ${CLAUDE_SKILL_DIR}/references/example-audit.md for complete PASS and FAIL examples.

</example_review>

<success_criteria>

A sound verdict has these properties:

• [ ] The verdict states exactly one overall determination: APPROVED or REJECTED
• [ ] Every applicable Python concern in the verdict table was judged, with no skipped concern hidden by an approval
• [ ] Each FAIL finding names the file, line, violated concern or rule, and concrete evidence
• [ ] Each NOT_APPLICABLE row explains why the concern does not apply; a missing or blocked required inspection produces FAIL
• [ ] The same repository state and audit scope can reproduce the verdict from the listed evidence

</success_criteria>