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 vagkaratzas/token-saviour --skill token-saviourgit clone --depth 1 https://github.com/vagkaratzas/token-saviourWrote 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/vagkaratzas/token-saviour/token-saviour)<a href="https://agentmods.dev/skills/vagkaratzas/token-saviour/token-saviour"><img src="https://agentmods.dev/badge/skills/vagkaratzas/token-saviour/token-saviour.svg" alt="Measured on agentmods" 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 | $0.00179 | $0.02769 |
| Opus 5 | $0.00089 | $0.01385 |
| Sonnet 5 | $0.00036 | $0.00554 |
| Haiku 4.5 | $0.00018 | $0.00277 |
Grade A, and why
token-saviour 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 7d 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 — 188 lines — stays where its author put it; the contents beside it link to each section on GitHub.
token-saviour: spend tokens where they matter
Reading whole files to answer a narrow question is the single most wasteful thing an agent does. On a benchmark over a ~30-file Python app, swapping whole-file reads for semantic retrieval cut total tokens ~66%; the other tools each own a narrower slice. This skill helps you reach for the right one before you blow the context budget — then get back to the actual task.
The mental model: token cost has four independent layers, and a different tool owns each.
Match the tool to the layer the task actually stresses. These winners were re-validated in a
9-tool benchmark (2026-07) against newer entrants (archex, Pare, lazy-cat) — every incumbent
held its layer; the runner-ups and the niches where they flip are in
references/tool_links.md.
| Layer | What it is | Tool | Don't bother with |
|---|---|---|---|
| Code-read input | Understanding code: symbols, callers, call paths, architecture | serena | rtk, caveman, Ponytail |
| Command-output input | Verbose stdout: tests, builds, git, grep, listings | rtk | the code-read tool |
| Generated prose output | Your own long, chatty replies / write-ups | caveman | the input tools, Ponytail |
| Generated code output | Implementations you write/edit | Ponytail | the input tools, caveman |
| — | Tiny/one-off work | plain Read/Grep/Bash | everything (overhead > benefit) |
First, check availability. These are optional third-party tools. Run the relevant
--help/--versiononce; if a tool isn't installed, seereferences/tool_links.mdfor install + verify commands, or fall back to the next-best option in its row (ultimately plain Read/Grep/Bash). Never pretend a tool ran — degrade gracefully.
Announce what you used. Whenever this skill drives a task, print a one-line tag naming the token-saving tools actually used, joined with
+— e.g.🪙 token-saviour: serena + rtk + caveman, or🪙 token-saviour: serena + ponytail. If you fell back to plain Read/Grep/Bash because nothing was installed or the task was trivial, say so:🪙 token-saviour: plain tools (fallback). List only tools you genuinely invoked, in layer order (serena → rtk → caveman / ponytail).
What ships with it
2 files 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.
- 7d ago First seen · 188 lines · 179 tokens per session scan A 9fdd1be38cb0
token-saviour is a skill published in the GitHub repository vagkaratzas/token-saviour (10 stars, last pushed 2mo ago), licensed MIT. It adds 179 tokens to every session and 2,769 once invoked, about $0.0009 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 2026-08-31.
Other skills, from other repositories
tokf-discover
Find missed token savings by scanning AI coding session files for commands that ran without tokf filtering.
calendar
Create, list, and manage calendar events / meetings from the terminal using whatever native CLI is installed (gcalcli for Google Calendar, khal for CalDAV, or AppleScript Calendar on macOS). Use when asked to "schedule a meeting", "create an event", "what's on my calendar", "am I free", "book".
Send and read email from the terminal using whatever native CLI is installed (himalaya, msmtp/mutt/neomutt on Linux, or AppleScript Mail on macOS). Use when asked to "send an email", "email someone", "check my inbox", "read my email", "reply to".
reminders
Create and list reminders / to-dos and time-based nudges using native tooling (AppleScript Reminders on macOS, or ChatCLI's own @scheduler for a timed message back to the user). Use when asked to "remind me", "set a reminder", "add a to-do", "don't let me forget".
send-message
Proactively send a message to the user (or a chat) over a connected messaging platform — Telegram, WhatsApp, Discord, Slack, or a generic webhook. Use when asked to "notify", "ping", "message", "tell", "send", "let me know on Telegram/WhatsApp", or to deliver a result to a chat.
notion
Read and write Notion — search pages/databases, query a database, create pages, append blocks — via the official Notion REST API and the user's own integration token. Use when asked to "add to Notion", "find in my Notion", "create a Notion page", "update my Notion database".