proofread-subtitles — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited proofread-subtitles (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.
Ask author for domain-specific terms engine likely misrenders. Use that list as primary reference for errors. Proceed without list when author declines. Flag uncertain cues by number instead.
Keep timestamps, cue numbers, and blank line between cues unchanged. Keep each cue's boundaries exactly as in source. Break any line over 42 characters into two within same cue. Split at clause boundary, comma, or conjunction. Allow at most two text lines per cue. Shorten wording rather than add third line.
Add punctuation where spoken phrasing requires it. Follow source language's rules. Capitalize first word of every sentence and every proper noun. End every sentence with terminal mark.
Fix obvious homophone and word misrecognitions when context is unambiguous. Preserve computer jargon in its conventional form. Use API, CLI, Git, Docker, JSON, HTTP, and Kubernetes as written. Fix typos and spacing errors without rewording sentence.
In Russian subtitles, write computer terms in English. Follow conventions of Russian tech speech. Write pull request instead of пулл реквест. Keep transliterations already standard in Russian.
Preserve speaker's wording verbatim. Keep filler words like uh, um, well, ну, and вот. Delete them only when author requests filler-free transcript. Omit speaker labels and sound descriptions. Add bracketed annotations only where source already uses them. Preserve file's existing encoding and line endings.
Leave original when unsure change preserves meaning. Flag cue number instead.
Write output to sibling file <original>-corrected.srt. Keep source file intact. Verify cue count matches source. Verify every cue number and timestamp matches too. Report short categorized summary of changes at end.
Correct this raw cue.
1
00:00:01,000 --> 00:00:03,000
lets push this to git hub tomorrowReturn this cue.
1
00:00:01,000 --> 00:00:03,000
Let's push this to GitHub tomorrow.~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.