fix: resolve ruff-check lint errors in non-template Python files

[TheSaiEaranti]

Fixes #218

Scope

ruff check (v0.15.0, the pinned pre-commit rev) reports 91 findings on main. They split cleanly in two:

	findings	files
code_templates/*.py	46 (44 invalid-syntax, 2 E402)	12	#214 — untouched here
everything else	45 (F541 35, F841 5, F401 4, F821 1)	12	this PR

Non-template findings are now 0. The 46 belonging to #214 are unchanged.

Two notes on how this differs from the issue description, which was written on 2026-07-06:

• The affected non-template file list has grown from 4 to 12 as the repo has gained scripts.
• E402 is no longer among the non-template codes — both remaining E402 findings are inside code_templates/, so they belong to pre-commit: 12 code_templates/*.py files fail Python-parse hooks (check-ast, ruff-format) #214. The non-template codes are now F541, F841, F401 and F821.

Mechanical fixes

Applied with ruff check --fix (safe fixes only; no --unsafe-fixes):

• F541 (35) — dropped the f prefix from f-strings with no placeholders.
• F401 (2 of 4) — removed typing.Union from validate_custom_metrics.py and INT from test_safe_query.py. I checked that INT is not reachable through that file's globals()-based fallback test discovery, which selects only callables named test_*; INT_INJECTIONS is a separate name and is still used.

Reviewed individually

tools/validate-urls.py — F821 undefined name httpx
This is a PEP 723 inline script that imports httpx lazily inside check_url and main, so the module-scope annotation client: "httpx.AsyncClient" had nothing to resolve against. Added a TYPE_CHECKING-guarded import rather than a noqa, which keeps the lazy-import behaviour intact — I verified the module still imports with httpx absent, and the annotation is preserved.

verify-traceability.py — F841 overall_pct
generate_html computed overall_total, overall_impl and overall_pct, but only rules_pct and reqs_pct are ever rendered. Removing just overall_pct would have left the other two unused, so all three go. This looks like an intended "overall" dashboard stat that was never wired into the HTML — worth a follow-up if you want that row displayed rather than deleted.

post_process_drawio.py — F841 actor_x2
Removed. Flagging what it hints at, since it may be worth its own issue: the overlap test on the next line is

if actor_x >= cloud_x and actor_x < cloud_x2:

which only detects an actor whose left edge is inside the AWS Cloud. An actor starting left of the cloud and extending into it is not detected. actor_x2 is exactly the term a full interval-overlap test would need, so this reads like an unfinished check. I have not changed that behaviour here — it is a layout change, not a lint fix, and should be your call.

Templates

The three *_source_template.py files are customer-facing starter templates ("INITIAL SUGGESTION ONLY - MUST BE CUSTOMIZED"), where from typing import Dict, Any, List, Optional # For type hints and the example variables are deliberate scaffolding for the reader. Stripping them would remove the thing the template exists to provide, so each gets a file-level directive instead, as the issue suggests:

# Starter template: the imports and example variables below are intentional
# scaffolding for the reader to build on, so unused-name rules do not apply.
# ruff: noqa: F401, F841

Verification

• ruff check . → non-template findings 0; remaining 46 are all code_templates/* (pre-commit: 12 code_templates/*.py files fail Python-parse hooks (check-ast, ruff-format) #214).
• All 12 changed files parse (ast.parse) and byte-compile.
• tools/evals/databases-on-aws/dsql/scripts/test_safe_query.py → 63 passed, 0 failed.
• validate-urls.py imports at module scope with httpx not installed.
• verify-traceability.py and post_process_drawio.py behave identically to main on import (the latter raises the same pre-existing defusedxml AttributeError in my environment both before and after).

Out of scope: ruff-format fails repo-wide on main — 25 files, including files this PR touches. That count is 25 before and 25 after, so nothing here adds to it. Running the formatter would turn this into an unrelated 25-file diff, so I left it alone; happy to file it separately if it is not already tracked under #213.

---

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of the project license.