Borrowing it
Nothing to install: this file belongs to dginev/latexml-oxide. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/dginev/latexml-oxide/main/.claude/skills/perf-check/SKILL.mdgit clone --depth 1 https://github.com/dginev/latexml-oxideWrote 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/dginev/latexml-oxide/perf-check)<a href="https://agentmods.dev/skills/dginev/latexml-oxide/perf-check"><img src="https://agentmods.dev/badge/skills/dginev/latexml-oxide/perf-check/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/dginev/latexml-oxide/perf-check"><img src="https://agentmods.dev/badge/skills/dginev/latexml-oxide/perf-check.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 | $0.00085 | $0.01323 |
| Opus 5.5 | $0.00034 | $0.00529 |
| Sonnet 5.5 | $0.00017 | $0.00265 |
| Haiku 4.5 | $0.00009 | $0.00132 |
Grade A, and why
perf-check 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 2mo 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 — 80 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Build profiles — use the right one (Cargo.toml)
| Profile | Command | Use |
|---|---|---|
test (default) |
cargo build / cargo test / cargo run |
All day-to-day dev + triage. Full debug info, debug-assertions, overflow-checks, fast incremental. Best diagnosability. |
ci |
cargo test --profile ci |
GitHub runner only (16 GB, fast compile). NOT what local dev should mimic. |
release |
cargo build --release |
Sandbox sweeps + Perl-parity measurement. Strong-optimized, thin-LTO. |
maxperf |
cargo build --profile maxperf |
Distribution artifact. Fat-LTO, CGU=1, panic=abort. Slowest build; smallest+fastest binary. |
Distribution recipe (what tools/make_release.sh uses): cargo build --no-default-features --features runtime-bindings --profile maxperf --bin latexml_oxide.
Settled questions — DO NOT re-open (measured, closed, tooling deleted)
- PGO / target-cpu (
x86-64-v3,native): NO measurable gain. Rigorous measurement (2026-06-21, disjoint train/measure split, best-of-3 interleaved): total speedup 0.999×, identical medians.maxperfis already at the compiler ceiling (fat-LTO + CGU=1 captures what PGO would unlock on thin-LTO), and the hot path (branchy catcode/macro dispatch) is neither SIMD-amenable nor statically branch-skewed. Ship the portablex86-64baseline. Do not re-attempt without a major engine-architecture change. (pgo-isa-no-gain-2026-06-21memory;docs/performance/PERFORMANCE.md.) - Startup floor (~161 ms): the ~50 ms dump-parse lever was declined.
Decomposed: proc init ~6 ms, bootstrap ~9 ms, latex dump load ~85 ms
(60% text-parse),
_constructs+digest+build+serialize ~80 ms. The clean lever (binary dump format / parallel-parse) is ~50 ms ≈ 10% of the 507 ms median — declined as too small to justify the dump-pipeline parity re-validation risk. Compiling dumps to Rust code (vs gzip-blob embed) cost +10 min compile for zero win, abandoned. Do not optimize the dump load. (docs/archive/STARTUP_COST_ANALYSIS_2026-06-21.md.)
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.
- 2mo ago First seen · 80 lines · 85 tokens per session scan A 072c91c6087d
perf-check is a skill published in the GitHub repository dginev/latexml-oxide (18 stars, last pushed yesterday), licensed CC0-1.0. It adds 85 tokens to every session and 1,323 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 2026-08-30.
perf-check compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| perf-checkdginev/latexml-oxide | Skill | 18 | yesterday | CC0-1.0 | A | 85 | Claude Code |
| inclusio-publishersebastienrousseau/inclusio | Skill | 0 | 2d ago | Apache-2.0 | A | 30 | Claude Code, Codex |
| Paper Writing Skills Indexaristoteleo/PantheonOS | Skill | 493 | yesterday | BSD-2-Clause | A | 76 | Claude Code, Codex |
| test-iterate-loopflonat/flonat-research | Skill | 149 | 3d ago | MIT | A | 55 | Claude Code |
| postmortemflonat/flonat-research | Skill | 149 | 3d ago | MIT | A | 40 | Claude Code |
| code-archaeologyflonat/flonat-research | Skill | 149 | 3d ago | MIT | A | 50 | Claude Code |
| latex-rescueCalix-L/awesome-latex-skills | Skill | 185 | 1mo ago | MIT | A | 41 | Codex |