voice-gate — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited voice-gate (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
This skill is a coach, not a gatekeeper. It runs the VOICE.md rubric over a draft and flags concerns. It does NOT block publish. The author reviews findings and decides whether each is a real issue or a false positive.
Why advisory: regex-based phrase detection can't reliably distinguish use from citation. "Leverage" the verb is fine in Cooper's writing; "leverage synergistic value" is not. Without an LLM judge, the rubric over-fires on legitimate prose. Calibrated to threshold 70 against Cooper's existing writing samples, but VISION.md (43.5) and the architecture audit (0) score below threshold because they cite the very phrases they're warning against.
What this catches reliably: clean LLM-default drafts. "We're leveraging AI to revolutionize..." scores ~70 even with the conservative weights — and a human reading the findings sees the three banned phrases highlighted, knows to rewrite.
What this misses: subtle drift, generic-but-not-cliché writing, content that's technically clean but voicelessly bland.
Before publishing any of:
Don't use for: postmortems Cooper writes himself, internal decisions docs, code comments, Telegram replies, vault notes, audit doc sections.
python3 ${REPO_ROOT}/tools/voice/score.py "$PATH_TO_DRAFT" "$SURFACE"Surface options: default, newsletter, brief, product, postmortem, telegram. Each has a different sentence-length target.
Output is JSON with:
score: 0-100passes: true if ≥70 (advisory threshold)stats: sentence count, paragraph count, avg sentence wordsfindings: per-category deductions (banned-phrases / sentence-length / specificity / passive-voice / filler-tokens)For each finding, the author considers:
The drafting skill (newsletter-weekly-publish, ship-log-draft, founder-letter-draft) calls voice-gate after generating the draft. If score < 70, the drafter prepends a revision note to the draft:
Voice-gate flagged this draft (score N/100). Consider: <list of top 3 findings>. Revise or override.
The author (LLM or Cooper) reads the note + findings and either revises or proceeds. Final approval is the author's, not the gate's.
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.