Borrowing it
Nothing to install: this file belongs to majiayu000/vibeguard. 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/majiayu000/vibeguard/main/.claude/skills/benchmark-regression-triage/SKILL.mdgit clone --depth 1 https://github.com/majiayu000/vibeguardWrote 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/majiayu000/vibeguard/benchmark-regression-triage)<a href="https://agentmods.dev/skills/majiayu000/vibeguard/benchmark-regression-triage"><img src="https://agentmods.dev/badge/skills/majiayu000/vibeguard/benchmark-regression-triage/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/majiayu000/vibeguard/benchmark-regression-triage"><img src="https://agentmods.dev/badge/skills/majiayu000/vibeguard/benchmark-regression-triage.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 | $0.00057 | $0.02332 |
| Opus 5 | $0.00028 | $0.01166 |
| Sonnet 5 | $0.00011 | $0.00466 |
| Haiku 4.5 | $0.00006 | $0.00233 |
Grade C, and why
benchmark-regression-triage scanned grade C with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 11d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Recursive force deletehighDestructive command
rm -rf with a variable or a broad path is one typo away from removing the wrong tree.
rm -rf "$OUT" How it starts
The opening of the file, as written. The whole thing — 255 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Benchmark Regression Triage
Overview
This is a repo-local maintainer skill for VibeGuard contributors. It is intentionally stored under .claude/skills/ so maintainers can use it while working in this repository, but it is not an installable product skill and must not be added to schemas/install-modules.json unless it is promoted into skills/ or workflows/ with a user-facing support contract.
This skill diagnoses VibeGuard hook latency regressions by comparing GitHub Actions bench-output artifacts across PR runs, merge runs, and mainline runs. It is meant for non-obvious cases where the current benchmark is under budget but slower than a previous design, such as a post-write-guard path losing its post-write-fast-check fast path after a runtime migration.
Treat CI artifacts as the trend source. Local benchmarks are useful for reproduction after a hypothesis exists, but they are not comparable to GitHub runner history because machine load, shell startup, cache state, and --runs count can dominate P95.
When to Activate
- A user asks why VibeGuard benchmark numbers are slower than before.
- A PR appears to change hook latency, benchmark output, or benchmark-action reporting.
- You need to compare recent PRs with older anchors such as PR #350 or a known fast-path implementation.
- A hook is still below the absolute latency budget but may have lost a faster design.
- You need a reusable workflow for downloading and comparing GitHub Actions
bench-outputartifacts.
Inputs
Collect these before drawing conclusions:
- Repository full name from
gh repo view. - Candidate PR numbers or commits, including one recent run and one older anchor.
- GitHub Actions run IDs tied to the exact PR head, merge commit, or main commit.
bench-output.jsonfrom each run, downloaded into separate directories.- The relevant budget contract from
docs/reference/hook-latency-contract.md.
Workflow
1. Search Existing Context
Search before adding a new hypothesis or artifact path:
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 First seen · 255 lines · 57 tokens per session scan C 3a900bb252b2
benchmark-regression-triage is a skill published in the GitHub repository majiayu000/vibeguard (41 stars, last pushed today), licensed MIT. It adds 57 tokens to every session and 2,332 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it C with 1 finding (recursive force delete). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
truecourse-fix
Fix TrueCourse violations that have suggested fixes.
truecourse-list
List TrueCourse violations found in this repository.
ring:searching-code
Forensic code search and analysis with optional Chain of Draft (CoD) ultra-concise mode. Five-phase methodology (clarification, planning, execution, analysis, synthesis) with severity assessment. Use for targeted investigation of specific patterns, bugs, or vulnerabilities. Skip for broad architecture mapping (use…
ring:using-assert
Using lib-observability/assert, Lerian's runtime assertion package, in two modes. Sweep Mode detects panic()/log.Fatal zero-panic violations, DIY invariant checks, hand-rolled domain predicates, and missing InitAssertionMetrics. Reference Mode catalogs the asserter lifecycle, domain predicates, observability trident…
ring:fixing-lint
Fixing lint to a clean state: runs the linter, groups reported issues into independent streams, and dispatches one parallel fixer agent per stream (ring:backend-go for Go, ring:general-purpose otherwise), iterating until clean. Use when a codebase has lint errors across multiple files. Skip for a single error (fix…
ring:test-driven-development
Enforcing the RED-GREEN-REFACTOR loop: write one failing test and watch it fail, write minimal code to pass, then refactor green. Use when starting implementation of a new feature or bugfix, or writing any new production code. Requires pasted failure output as proof of RED; code written before its test must be…