Transcript Fixer

What it does

• Stage 1 – Dictionary pass: instantly and freely applies accumulated rules from ~/.transcript-fixer/corrections.db to recurring errors (product names, people, frequent homophones). Safe mode is the default, so risky substitutions are deferred to a *_needs_review.md sidecar instead of being applied silently.
• Stage 2 – Native AI correction: with no external API key, Claude reads the Stage 1 output and fixes proper nouns, domain jargon and context-dependent homophones the dictionary has never seen. This is the primary path and does most of the real work on clean ASR.
• Compounding learning: confirmed fixes go into a domain dictionary via --add, repeated patterns are surfaced by --review-learned and promoted with --approve, and uncertain items live in a persistent review queue with one-keystroke verdicts.
• Guardrails: --dry-run previews, always-on change reports, --report-false-positive, rule auditing with --audit, corpus evidence with --probe, and a fail-closed anchor guard for queued edits.
• Automation fallback: agent-less CI can complete through a Stage 3 GLM API batch mode.

Who it's for

• Researchers, PMs and note-takers who regularly clean up meeting, interview or lecture transcripts
• Anyone turning Whisper / Otter / built-in meeting-tool output into publishable documents
• Teams whose internal product names and colleague names get mangled the same way every time
• Developers wiring transcript ingest pipelines that need a stable JSON status contract

Examples

1. Weekly meeting notes: run --stage 1 --domain myproject --apply-domain, let Claude read the result and fix the rest, then --add the recurring mistakes so next week they auto-fix.
2. Batch lecture transcripts: loop Stage 1 over a folder, read the --json line for applied/deferred, and only hand-review the deferred items.
3. Name normalization: register frequent people's ASR variants in the people roster so every future transcript spells them correctly without new rules.

Install the skill I found on Claude Skill Mart.
Copy the daymade-audio/transcript-fixer folder from the GitHub repo daymade/claude-code-skills into my ~/.claude/skills/transcript-fixer/.
When it's done, tell me in one line what this skill can do.