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 smixs/iva-agent --skill watchgit clone --depth 1 https://github.com/smixs/iva-agentWrote 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/smixs/iva-agent/watch)<a href="https://agentmods.dev/skills/smixs/iva-agent/watch"><img src="https://agentmods.dev/badge/skills/smixs/iva-agent/watch/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/smixs/iva-agent/watch"><img src="https://agentmods.dev/badge/skills/smixs/iva-agent/watch.svg" alt="Reviewed on agentmods" width="80" height="20"></a>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 | $0.00176 | $0.03423 |
| Opus 5.5 | $0.00070 | $0.01369 |
| Sonnet 5.5 | $0.00035 | $0.00685 |
| Haiku 4.5 | $0.00018 | $0.00342 |
Grade A, and why
watch 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 yesterday.
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 — 199 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Watch — telling the owner what they missed
Code checks the owner's Telegram and mail once an hour without you and wakes you only when something new has waited long enough (or an urgent sender wrote). Your job is the judgement: is it worth the owner's attention, and what is the next step.
Shared procedure for Watch, Brief and Insight
- For an agreed responsibility, use the subject, available sources, notify condition
and end condition. In chat, reuse the owner's details; ask one material missing
detail. Save confirmed behaviour in Owner rules through
write_file; facts go in Card. Existing confirmation is enough. Use current Watch/Brief turns. Every Routine firing reaches the owner; remove it only at their request. - Apply Owner rules first;
memory_searchthe subject andinsightCards before choosing. Skip declined proposals, including rewordings, unless asked again. Check CORE goals against tasks and newer Card facts. CORE goal advice needs an active CORE goal and a useful change; stop that advice when the end condition is met. - When a message is warranted, prepare a short reply draft or checked choices from the material read, within existing turn bounds. Name missing facts. Keep preparation in the same message. The owner sends replies; installation and repairs follow their existing approval rules. For each substantial item, give its source and why now, such as a new reply or due date. Use a source link, or its name and date when a link is unavailable.
- In chat, keep a single rejection on the proposal's Card, or one factual Card with subject and date; confirm its scope in one short line. Save a topic-wide rule in Owner rules only after explicit confirmation. Keep Insight's Card fields and archive step. Silence means no known reaction; "Got it" is one event.
- Apply this judgement with each turn's delivery rules. Alert gets its message, morning Brief always answers, Signal and Reminder deliver.
In a Watch turn
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.
- yesterday Changed · +24 lines · +22 tokens per session 1dd973e11375
- 7d ago Changed · +65 lines · +21 tokens per session bcc0c1ab2512
- 10d ago First seen · 110 lines · 133 tokens per session scan A ec2e0da73227
watch is a skill published in the GitHub repository smixs/iva-agent (240 stars, last pushed today), licensed MIT. It adds 176 tokens to every session and 3,423 once invoked, about $0.0007 per session on Opus 5.5. 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 2026-10-06.
Other skills, from other repositories
task_automation
Design effective automated workflows using scheduled tasks, prompt chaining, and delivery channels.
himalaya
Himalaya v2 CLI for reading and managing email through IMAP, JMAP, Gmail, Microsoft Graph, Maildir, M2dir, and SMTP send operations.
slacrawl
Archive and search Slack workspace messages, threads, and channels via the slacrawl CLI. Supports API sync, Slack export ZIP import, local desktop cache import, and full-text search.
birdclaw
Local-first Twitter/X workspace for archiving, searching, and triaging tweets, DMs, likes, bookmarks, and mentions. Includes a web UI for inbox triage and AI scoring.
apple-reminders
Manage Apple Reminders via remindctl CLI (list, add, complete, delete).
imsg
Use imsg to read and send iMessage/SMS via macOS Messages.app.