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 Goldziher/poly --skill poly-tiers-and-scopegit clone --depth 1 https://github.com/Goldziher/polyWrote 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/goldziher/poly/poly-tiers-and-scope)<a href="https://agentmods.dev/skills/goldziher/poly/poly-tiers-and-scope"><img src="https://agentmods.dev/badge/skills/goldziher/poly/poly-tiers-and-scope/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/goldziher/poly/poly-tiers-and-scope"><img src="https://agentmods.dev/badge/skills/goldziher/poly/poly-tiers-and-scope.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 | $0.00061 | $0.02099 |
| Opus 5.5 | $0.00024 | $0.00840 |
| Sonnet 5.5 | $0.00012 | $0.00420 |
| Haiku 4.5 | $0.00006 | $0.00210 |
Grade A, and why
poly-tiers-and-scope 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 11d 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 — 129 lines — stays where its author put it; the contents beside it link to each section on GitHub.
poly Tiers and Scope
Coverage tiers inside the per-file pass
Five mechanisms decide what actually inspects a file (crates/poly-core/src/registry.rs):
- Tier-1 native crate backends — one in-process Rust crate registered for specific languages: ruff, oxc, taplo, rumdl, sqruff, yaml, malva (+ biome for CSS/SCSS), markup_fmt, mago, rubyfmt, graphql, nixfmt, hcl, dockerfile, dotenv, ini.
- Tier-2 tree-sitter generic tier (ADR 0004) — the catch-all for every language with no native backend. Structural reindent for brace-family grammars, whitespace normalization otherwise, and a leave-untouched list where whitespace is significant. This is the coverage mechanism, not a fallback to avoid — but it is explicitly not gofmt/rustfmt parity.
- Native-toolchain tier (ADR 0014,
engines/native_tool/) — the language's canonical first-party CLI, run per file over stdin/stdout.rustfmt,gofmtandshellcheckare on by default whenever the tool is found on PATH;zig fmt,shfmt,google-java-format,ktfmt,styler,swift-format,dart format, andgleam formatare opt-in via[fmt.<lang>.<tool>] enabled = true/[lint.<lang>.<tool>] enabled = true. Absent tool ⇒ the language falls through to tier 2 with an info-level notice, never an error. quality, the code-quality metric tier (ADR 0027) — cross-cutting, appended to every language: file/function/type length, nesting depth, cyclomatic complexity, parameter count, andlazy-ignore. Warning severity, on by default, configured under[lint.quality]. The structural rules need a verified node-kind table and cover Python, Rust, Go, JavaScript, TypeScript, TSX, Java, Kotlin, C, C++, C#, and Ruby;magic-numberandlaw-of-demeterare opt-in. A per-language deferral table keeps it from duplicating a tier-1 rule (ruff'sC901/PLR0913, oxlint'smax-depth/max-params).- The built-in ast-grep rule pack (ADR 0029) — 26 rules across 9 languages (C#, Elixir,
Go, Java, Kotlin, Python, Ruby, Rust, Swift) embedded in the binary and on by default; 13
of them ship
severity: off. It sits beneath[rules] dirs(default.poly/rules), so a user rule with the sameidreplaces a pack rule.[rules] builtin = falsedisables the pack wholesale.
What ships with it
1 file 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.
- 11d ago Changed · +4 lines 4dfc402624f6
- 14d ago Changed · +6 lines bd7152a456c8
- 1mo ago Changed · +15 lines a30db3e06584
- 1mo ago First seen · 104 lines · 61 tokens per session scan A 2106f4ce46dd
poly-tiers-and-scope is a skill published in the GitHub repository Goldziher/poly (20 stars, last pushed 6d ago), licensed MIT. It adds 61 tokens to every session and 2,099 once invoked, about $0.0002 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 2026-08-31.
Other skills, from other repositories
roam
Codebase comprehension via roam-code CLI. Use when exploring codebases, planning modifications, debugging failures, assessing PR risk, or checking architecture health. Triggers on: understanding project structure, pre-change safety checks, finding symbols/files, blast radius analysis, affected tests, health scoring…
audit
Fast quality gate — detects stack, runs static analysis, checks cross-layer consistency, fixes issues. Run between every pipeline phase, before every commit push, after every feature batch. Lighter than /analyze but catches the same critical issues.
logic-lens
AI-powered Claude Code skill that performs deep code review using formal logic and reasoning frameworks to detect bugs, anti-patterns, and security risks beyond what linters catch.
argot-refresh
Refresh Argot's committed fit snapshot safely — first diagnose why maintenance is recommended, re-audit corpus scope and structural path changes, review stale mutes and policy entries with the user, then fit locally, verify, and prepare the reviewed .argot/ update. Use when argot status, argot check, MCP, or CI…
ph-debug
(PH) Use for a reproducible failure or unexpected behavior; reproduce, isolate, and verify the cause.
ph-lsp-setup
(PH) Use only when the user explicitly requests language-server setup or diagnostics.