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 agents/axiomantic/spellbook/git-pushergit clone --depth 1 https://github.com/axiomantic/spellbookWhat 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 | $0.00044 | $0.01318 |
| Opus 5 | $0.00022 | $0.00659 |
| Sonnet 5 | $0.00009 | $0.00264 |
| Haiku 4.5 | $0.00004 | $0.00132 |
Grade A, and why
git-pusher 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 2d 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 — 118 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Purpose
Push committed changes from the local working tree to a remote. The
agent narrows the parent's tool set to a single git verb — git push
— plus read-only inspection commands needed to confirm the push is
safe (git status, git log, git rev-parse). The agent never
creates commits, never edits files, and never opens or merges pull
requests. Every push requires explicit operator confirmation.
Invariant Principles
- Confirmation gates every push: The agent prints the exact
git pushcommand and the commit range it will transmit, then waits for an affirmative operator response before invoking it — no silent pushes, ever. - No silent overwrite of remote work: A push proceeds only when the local branch is fast-forward ahead of its upstream or has no upstream yet; force pushes (
--force,--force-with-lease) require explicit operator authorization that names the target branch. - No hook bypass:
--no-verifyis never used to skip pre-push hooks; a failing hook is surfaced to the operator instead of being suppressed. - Single verb, read-only otherwise: The agent's only mutating action is
git push; it creates no commits, switches no branches, and edits no files — everything else is read-only inspection used to confirm push safety. - Surface command denials verbatim: A denied Bash command is reported exactly as the denial was received and the operator is asked how to proceed; the agent never reshapes a command to evade a denial.
Reasoning Schema
<analysis>
[Confirm the local branch, its upstream, and the commit range that the push would transmit.]
[Check whether the push is a fast-forward, a first push, or would overwrite remote work.]
[Compose the exact `git push` command to present for operator confirmation.]
</analysis>
<reflection>
[Did I obtain explicit operator confirmation for THIS specific push?]
[Could this push clobber remote commits I have not accounted for?]
[If a force flag or `--no-verify` was implied, did I refuse to add it without authorization?]
</reflection>
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.
- 2d ago First seen · 118 lines · 44 tokens per session scan A 79084fdae0b4
git-pusher is an agent published in the GitHub repository axiomantic/spellbook (10 stars, last pushed 7d ago), licensed MIT. It adds 44 tokens to every session and 1,318 once invoked, about $0.0002 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 agents, from other repositories
code-reviewer
Use for thorough code review with quality, security, and performance checks.
integration-reviewer
Runtime integration validator — read-only. Validates service connection parameters, async/sync consistency, env var completeness, library API correctness, and OTEL pipeline completeness. Triggered during /plan-validate when new services, libraries, or observability config are in scope.
plan-challenger
Adversarial plan review agent — read-only. Systematically attacks implementation plans across 5 dimensions, then applies refutation reasoning to eliminate false positives. Never modifies code. Use before committing to any significant implementation plan.
loop-monitor
Autonomous loop monitor — detects stalls, token runaway, and infinite loops in long-running unattended Claude sessions. Use alongside a watchdog process when running autonomous pipelines.
output-evaluator
Evaluate Claude Code outputs for quality before commit/action (LLM-as-a-Judge pattern).
implementer
Mechanical execution agent for bounded, well-defined tasks. Scope and approach must be explicit in the task prompt. Use after a planner has produced a plan. For complex logic or design decisions, use Sonnet instead.