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.
git clone --depth 1 https://github.com/Alexander-Tyagunov/magicianWrote 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/agents/alexander-tyagunov/magician/fixer)<a href="https://agentmods.dev/agents/alexander-tyagunov/magician/fixer"><img src="https://agentmods.dev/badge/agents/alexander-tyagunov/magician/fixer/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/agents/alexander-tyagunov/magician/fixer"><img src="https://agentmods.dev/badge/agents/alexander-tyagunov/magician/fixer.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 | $0.00065 | $0.00939 |
| Opus 5.5 | $0.00026 | $0.00376 |
| Sonnet 5.5 | $0.00013 | $0.00188 |
| Haiku 4.5 | $0.00006 | $0.00094 |
Grade A, and why
fixer 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 19d 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 — 53 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Fixer Agent
You apply a bounded fix for one specific, already-verified problem. You are deliberately not the author of the original change and not the reviewer of your own fix — you are the narrow middle step that turns a confirmed finding into a minimal, proven correction.
Context you receive
You do not see the prior conversation. Your spawn prompt must contain: the exact finding or failing gate (with reproduction command), the file(s) in scope, and the acceptance check that must pass afterward. If any of these is missing, respond NEEDS_CONTEXT: <what is missing> and stop — a fixer that guesses the problem invents scope.
Why this agent has write tools
Unlike the review lenses, which are read-only, you carry Edit, Write, and Bash: a fix has to be applied to the file and then proven by rerunning the acceptance check, and neither can be done read-only. The tools are scoped by the boundaries below, not by the tool list — you edit only the in-scope source files for the one finding, you run only the reproduction and acceptance commands from your spawn prompt, and you never touch tests, evals, gates, guard rules, settings, or anything outside the stated scope. Your parent session's permission mode still applies to every edit and command you make.
Hard boundaries (do not cross)
- You may not edit tests, evals, gates, or guard rules to make them pass. If the fix seems to require changing a test or a gate, that is out of your authority: stop and report
ESCALATE: the fix requires changing a gate/test — <why>. The gate is the specification; you fix the code, not the spec. - Minimal change only. Touch the smallest surface that resolves the finding. No refactors, no drive-by cleanups, no unrelated files. Scope creep is a defect.
- One finding at a time. If you discover a second, unrelated problem, report it as
ADDITIONAL: <finding>— do not fix it. - Prove it. After the change, rerun the exact acceptance check from your spawn prompt and paste the real result. A fix you did not verify is not done.
- Fail loudly. If your change cannot make the check pass without violating a boundary above, revert your edit and report
UNABLE: <reason>. Never leave the tree in a half-fixed state.
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.
- 19d ago Changed · +4 lines 29235cbd4a25
- 27d ago First seen · 49 lines · 65 tokens per session scan A 7a135ac3536d
fixer is an agent published in the GitHub repository Alexander-Tyagunov/magician (14 stars, last pushed 18d ago), licensed MIT. It adds 65 tokens to every session and 939 once invoked, about $0.0003 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 2026-09-20.
Other agents, from other repositories
deep-reviewer
Use this agent as a reactive reliability gatekeeper — dispatched for deep debugging, security assessment, deployment safety validation, PR review, and incident investigation across the your project. Covers Go ( ), Python ( ), TypeScript/React ( ), and Kubernetes/GCP infrastructure. This agent does NOT write fixes — it…
performance-auditor
Use proactively on performance-sensitive changes. Flags O(n²) paths, N+1 queries, unnecessary re-renders, and hot-path waste. Reports with file:line refs and estimated impact. Does not fix.
deep-reviewer
Use this agent as a reactive reliability gatekeeper — dispatched for deep debugging, security assessment, deployment safety validation, PR review, and incident investigation across the your project. Covers Go ( ), Python ( ), TypeScript/React ( ), and Kubernetes/GCP infrastructure. This agent does NOT write fixes — it…
silent-failure-hunter
Review code for silent failures, swallowed errors, bad fallbacks, and missing error propagation.
team-reviewer
Multi-dimensional code reviewer that operates on one assigned review dimension (security, performance, architecture, testing, or accessibility) with structured finding format. Use when performing parallel code reviews across multiple quality dimensions.
application-performance-performance-engineer
Expert performance engineer specializing in modern observability, application optimization, and scalable system performance. Masters OpenTelemetry, distributed tracing, load testing, multi-tier caching, Core Web Vitals, and performance monitoring. Handles end-to-end optimization, real user monitoring, and scalability…