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 agentmods add commands/kubalajiri/claude-squirrel/squirrel-taskgit clone --depth 1 https://github.com/kubalajiri/claude-squirrelWrote 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/commands/kubalajiri/claude-squirrel/squirrel-task)<a href="https://agentmods.dev/commands/kubalajiri/claude-squirrel/squirrel-task"><img src="https://agentmods.dev/badge/commands/kubalajiri/claude-squirrel/squirrel-task.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 | $0.00021 | $0.00822 |
| Opus 5 | $0.00010 | $0.00411 |
| Sonnet 5 | $0.00004 | $0.00164 |
| Haiku 4.5 | $0.00002 | $0.00082 |
Grade A, and why
squirrel-task scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 5d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 56 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Run this exact command with the Bash tool and show the user its output verbatim:
sh "${CLAUDE_PLUGIN_ROOT}/scripts/squirrel.sh" --task "$ARGUMENTS"
How to build a watcher, in order
- Read the vocabulary first —
/squirrel-metrics. Never write a condition from memory of a metric name; the installed version is the only one that can accept or refuse it. - Enable the action if it is not already —
/squirrel-task-actions notify on. A task whose action is off is refused, not created dormant. - Create it —
add <id> when='<condition>' do=<verb> text='...' cooldown=<seconds>. - Dry-run it —
/squirrel-task test <id>. You cannot see the user's bar, so this is how you check your own work before telling them it works. It reports whether the condition holds right now, and if not, the metric's actual value. - Show the user what you made, in plain words: what it watches, what it will do, how often at most, and how to stop it. A watcher that acts on their machine is not something to create silently.
The fields
| Field | Meaning |
|---|---|
when |
<metric> <op> <value>, or a bare <metric>. Same grammar as /squirrel-rules. |
do |
notify, run, ack or snooze. Each needs its own opt-in. |
text |
the exact message a notify sends — squirrel never composes one of its own |
target |
idle (any idle session) or a session name from claude agents --json |
cooldown |
seconds between firings. Minimum 60, default 1800. A 0 is silently raised. |
warn |
seconds of visible countdown before it fires (0 fires at once) |
cancel |
whether that countdown can be clicked to dismiss this firing |
Things worth telling the user
- Dismissing a nudge buys them the cooldown, not a second. Cancelling suppresses that firing and starts the cooldown; the task itself is untouched and fires again after it.
- A nudge costs a real request on their own account, capped by
nudgeMaxPerHour. - Unknown never fires. A condition over a metric with no value does not match,
!=included. - Squirrel only ever queues a job; a reader acts on it. Nothing acts from the render.
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.
- 5d ago First seen · 56 lines · 21 tokens per session scan A 0854a4add06a
squirrel-task is a command published in the GitHub repository kubalajiri/claude-squirrel (2 stars, last pushed 10d ago), licensed MIT. It adds 21 tokens to every session and 822 once invoked, about $0.0001 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 commands, from other repositories
status
Show claude-bionify's current settings.
set
Change a claude-bionify setting, e.g. "fixation 0.7" or "boundary syllable".
off
Turn claude-bionify off so Claude's replies render normally.
on
Turn claude-bionify on so Claude's replies are bolded.
reset
Clear all live claude-bionify overrides, back to your configured defaults.
toggle
Flip claude-bionify on or off.