fix: strip zero-width space (U+200B) from test_results.csv - #5
Conversation
Renovate raised a repository problem on this repo: "Hidden Unicode characters
have been discovered in file(s) in your repository … they could be an attempt to
'smuggle' text into your codebase, or used to confuse tools like Renovate or
Large Language Models (LLMs)".
Traced to a single U+200B at line 2865, trailing the end of an Entra
recommendation description ("…enhancing overall security posture.<U+200B>") —
a copy-paste artifact from the Microsoft source text, not anything hostile.
It is the only invisible character in the repository.
Removed binary-safely so the file's CRLF line endings are untouched: the diff
is one line, 340619 -> 340616 bytes.
Super-linter summary
Super-linter detected linting errors For more information, see the GitHub Actions workflow run Powered by Super-linter SPELL_CODESPELL |
Touching test_results.csv makes super-linter lint it for the first time, and
codespell then fails the SPELL_CODESPELL status with ~19 findings
("targetting", "unsecure", "appplications", "assiged", "Pretected", "usres", …).
All of them are pre-existing and all are Microsoft's own Entra recommendation
text reproduced verbatim in this generated Maester artifact — 8 are already on
main, and none appears in this branch's diff. Correcting them would falsify a
recorded test result rather than fix prose we own.
Skip the file, not the words, so codespell stays meaningful everywhere else
including future hand-written docs. Verified locally: codespell reports findings
without this config and exits clean with it.
Note this status was silently failing already: soft-launch gives the lint JOB
continue-on-error (main's run shows green), but super-linter posts a separate
commit status per linter which escapes it — the documented trap.
|
Pushed The
Touching the file simply makes super-linter lint it for the first time. Correcting Microsoft's spelling inside a recorded test result would falsify the artifact, so the fix is to skip the file (not the words) — codespell stays active for everything else, including future hand-written docs. Verified locally: findings without the config, exit 0 with it. Worth knowing: this status was already failing silently. |
Super-linter summary
All files and directories linted successfully For more information, see the GitHub Actions workflow run Powered by Super-linter |
Clears the Mend repository problem on this repo:
What it was — benign, but worth removing
A single U+200B (zero-width space) at
test_results.csv:2865, trailing the end of an Entra recommendation description:That is a copy-paste artifact from the Microsoft source text, not anything hostile. A full scan (zero-width, BOM, bidi overrides, soft hyphen, word-joiner ranges) found no other invisible characters anywhere in the repository.
Renovate's warning is still the right call — an invisible character in a committed file is exactly the shape of a text-smuggling attempt, and it is worth keeping the detector meaningful rather than muting it.
How
Removed binary-safely so the file's CRLF line endings survive — a naive text-mode rewrite re-wrote all 4,125 lines. The diff is one line; 340,619 → 340,616 bytes.
Worth a separate decision:
test_results.csvis a 4,127-line Maester run artifact committed to the repo (added in #1) and nothing in-repo generates it. If it is meant to be a snapshot, fine; if not, it may belong in.gitignore/ a workflow artifact instead. Not changed here.