Minutes is an open-source, local-first meeting recorder that transcribes meetings, calls, and voice memos on the device and stores the results as Markdown files. It is for people who want their meeting records available to Claude Code, Codex, Cursor, and other MCP clients without uploading them. The catalogue add-ons extend agent workflows for querying and using those local meeting files.
Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx skills add silverstein/minutes --skill minutes-cleanupgit clone --depth 1 https://github.com/silverstein/minutesWrote this? Show the measurements
A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.
[](https://agentmods.dev/skills/silverstein/minutes/minutes-cleanup)<a href="https://agentmods.dev/skills/silverstein/minutes/minutes-cleanup"><img src="https://agentmods.dev/badge/skills/silverstein/minutes/minutes-cleanup/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/skills/silverstein/minutes/minutes-cleanup"><img src="https://agentmods.dev/badge/skills/silverstein/minutes/minutes-cleanup.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 4 findings, up to high
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- high Tool Misuse · line 77 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
- high Tool Misuse · line 83 Tool parameters are crafted to achieve unintended or unsafe behavior. Parameter abuse can bypass intended safety checks (e.g. shell=True, --force, dangerous glob patterns).Fix: Validate all tool parameters against an allowlist. Reject dangerous parameter values (shell=True, --force, -rf /) and use safe defaults.
- medium Rogue Agent · line 54 Skill establishes unauthorized persistence across sessions via cron jobs, startup scripts, or state files. Session persistence allows an attacker to maintain access beyond the current interaction.Fix: Remove any persistence mechanisms (cron jobs, startup scripts, state files). Skills should not maintain state across sessions without explicit user consent.
- medium Excessive Agency · line 88 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
What it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5.1 | $0.00060 | $0.00801 |
| Opus 5 | $0.00030 | $0.00400 |
| Sonnet 5 | $0.00012 | $0.00160 |
| Haiku 4.5 | $0.00006 | $0.00080 |
Grade A, and why
minutes-cleanup scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 96 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/minutes-cleanup
Help the user manage disk space and organize old recordings. Minutes is transcript-first: markdown notes and structured memory are durable, while raw audio is a temporary recovery/reprocessing layer unless pinned.
Check current usage
minutes storage
minutes storage --json
Present this to the user before taking any action.
Common cleanup tasks
Preview raw-audio cleanup
After transcription, the original audio files are no longer needed for search or
recap. They only matter if you want to re-transcribe with a better model or
debug recovery. Cleanup is preview-only unless --apply is passed.
# Use the configured retention policy
minutes cleanup
# Try a shorter successful-audio window
minutes cleanup --older-than 14d
# Machine-readable preview for agents
minutes cleanup --json
Delete raw audio candidates
Only apply cleanup after showing the preview and getting explicit confirmation.
minutes cleanup --apply
minutes cleanup --older-than 14d --apply
Archive old meetings
Move meetings older than N days to an archive folder:
mkdir -p ~/meetings/archive
# Find meetings older than 90 days
find ~/meetings -maxdepth 1 -name "*.md" -mtime +90
# Move them (confirm with user first)
find ~/meetings -maxdepth 1 -name "*.md" -mtime +90 -exec mv {} ~/meetings/archive/ \;
Archived meetings won't appear in minutes list or minutes search (which only scans ~/meetings/), but they're still on disk if needed.
Clean up processed voice memos
The watcher moves originals to ~/meetings/memos/processed/ after transcription:
du -sh ~/meetings/memos/processed/ 2>/dev/null
Clean up stale state
# Remove stale PID file
rm -f ~/.minutes/recording.pid
# Clean old logs (keep last 7 days)
find ~/.minutes/logs -name "*.log" -mtime +7 -delete 2>/dev/null
# Remove last-result.json (transient)
rm -f ~/.minutes/last-result.json
Gotchas
- Never delete
.mdfiles without asking — These are the transcripts. They're small and contain the actual value. WAV files are the space hogs. - Prefer
minutes cleanupover rawfind -delete— The CLI understands pinned audio and sidecar stems. - Archived meetings are invisible to search —
minutes searchonly walks~/meetings/and~/meetings/memos/. If you need archived meetings searchable, configure QMD to index~/meetings/archive/too. - Audio deletion is irreversible — If the user might want to re-transcribe with a better model later, suggest pinning important recordings and only deleting old unpinned candidates.
- Pin exceptions in frontmatter — Add
audio_retention: pinnedto a meeting to keep its raw audio out of cleanup candidates. - Audio is ~10 MB/minute, transcripts are ~1 KB/minute — Deleting audio saves 99%+ of space while keeping all searchable content.
- iCloud sync caveat — If
~/meetings/is in an iCloud-synced folder, deleted files go to "Recently Deleted" and still count against storage for 30 days.
What ships with it
1 file beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 11d ago First seen · 96 lines · 60 tokens per session scan A ad7ee5edd1d1
minutes-cleanup is a skill published in the GitHub repository silverstein/minutes (1,473 stars, last pushed today), licensed MIT. It adds 60 tokens to every session and 801 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
scheduling
Set up recurring or scheduled work — anything the user wants run on a schedule, repeatedly, or at a future time ("every morning", "each weekday at 9am", "daily digest", "weekly report", "remind me", "run this on a cron"). Teaches how to author a durable Vexa Scheduler routine that survives restarts. Use this instead…
steno
Make the agent aware of the user's Steno meeting notes and able to pull them in on request. Steno (https://github.com/stenolabs/stenoai) records, transcribes, and summarizes meetings entirely on-device, storing each as a Markdown summary plus a transcript. This skill locates that store cross-platform (macOS, Windows…
vexa-meetings
Work with the user's meetings through the Vexa MCP tools — send a bot to a call, follow the transcript while it runs, and write meetings into their notes (Obsidian, Notion, plain markdown) with speakers and action items. Use whenever the user mentions a meeting, a call, a transcript, meeting notes, or asks what was…
meeting_notes
Structure raw meeting notes into actionable minutes with follow-ups.
focus-and-priority-management
Delivers daily scheduling habits, meeting hygiene, attention management, and boundary-setting techniques for leaders whose calendars have become their strategy by default. Use when urgent tasks crowd out important ones, when meeting load exceeds productive work time, when a leader cannot recall their last hour of…
granola-to-steno
Sync Granola meeting notes into Steno's file-based store. Reads recent meetings from the Granola MCP (titles, dates, participants, AI summaries, verbatim transcripts) and regenerates Steno's output/summary.md and transcripts/transcript.txt files in a target folder (e.g. an iCloud Drive directory). The operation is…