qbcheck

qbcheck is a skill for Claude Code from 3awny/qship. It costs 73 tokens per session (3,945 once invoked), scanned A, original, MIT.

An adversarial check of reported bug findings that tests whether each one is real, reproducible, and correctly explained. It filters out false positives and overstated problems.

In plain words
What is it for?
Use it after a bug hunt or code review to validate reported bugs, confirm their causes, reject unsupported findings, and decide which issues are worth fixing.
Why use it?
It prevents developers from spending time fixing issues that are not genuine while retaining findings supported by the code. Each finding must include a clear claim and location.

Skill for Claude Code

Written for Claude Code: shipped in a Claude Code plugin. Also seen: reads .claude/ paths; mentions subagents; mentions Claude Code.

Part of the qship plugin — 22 skills, 1 agent, 1 hook shipped together

Install

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.

agentmods
npx agentmods add skills/3awny/qship/qbcheck
Any agent
npx skills add 3awny/qship --skill qbcheck
Clone the repo
git clone --depth 1 https://github.com/3awny/qship

Made for: Claude Code.

Or install qship, the plugin that ships this one along with the rest of its 22 skills, 1 agent, 1 hook.

Wrote 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.

agentmods badge for qbcheck

README.md
[![agentmods](https://agentmods.dev/badge/skills/3awny/qship/qbcheck.svg)](https://agentmods.dev/skills/3awny/qship/qbcheck)
Your own site
<a href="https://agentmods.dev/skills/3awny/qship/qbcheck"><img src="https://agentmods.dev/badge/skills/3awny/qship/qbcheck.svg" alt="Measured on agentmods" height="20"></a>
Per session 73 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 3,945 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00073 $0.03945
Opus 5 $0.00036 $0.01972
Sonnet 5 $0.00015 $0.00789
Haiku 4.5 $0.00007 $0.00394

Measured 5d ago against content hash 6b66eb85e3a1, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

qbcheck 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 5d 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.

templates/skills/qbcheck/SKILL.md · 228 lines

How it starts

The opening of the file, as written. The whole thing — 228 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Bug Review Validator

You are a skeptical senior engineer validating a list of bug findings. Your job is to separate real bugs worth fixing now from false positives, overstated severities, and overthinking — without flipping into the opposite failure mode of rejecting real bugs.

The goal is accuracy, not a low or high bug count. A typical pass rejects roughly 30–50% of automated findings, but the right number is whatever the code actually says.

Inputs

This skill takes a list of bug findings as input. The findings can come from anywhere:

  • /qbug output (most common)
  • A code-review subagent's report
  • PR review comments
  • A manual list the user pasted

What you need for each finding:

  • Claim — what the finding says is wrong
  • Location — file path + line number, or enough context to find it
  • Source — which tool/agent flagged it (e.g. silent-failure-hunter, security-scanner, "GitHub PR comment from X"). Use this as one signal, not a vote count.

If a finding is missing the location or claim, ask the user before guessing. Don't validate something you can't pin down.

What you produce

Two things, both required:

  1. Per-finding writeup — verdict + reasoning, in the format below.
  2. Final summary — human table + machine-readable filtered list (the bugs worth acting on, in order).

The machine-readable list is what downstream pipelines (e.g. /qship fix step) consume. Keep it clean.

Pre-flight: memory of past decisions (graceful — skip if unavailable)

Before validating, scan Claude Code's native auto-memory for past qbcheck verdicts on similar findings — this prevents re-litigating identical FP/TP judgments and reduces drift across runs. Per Memorisable Prompting. Defer to qmemory's rules; this section only describes the read + write shape qbcheck uses.

Read. The current project's memory directory is referenced in the system prompt's auto-memory section (typically ~/.claude/projects/<project-slug>/memory/) with MEMORY.md as the index. Skim MEMORY.md for existing entries that look like prior qbcheck verdicts — usually feedback_qbcheck_*.md (or anything tagged with the file/symbol you're validating). If you find ≤5 relevant entries, read them and use them as few-shot examples in your reasoning — "this finding is structurally similar to the one we marked False positive in feedback_qbcheck_X.md, which turned out wrong because…". Memory calibrates the verdict; it does not determine it.

Read the full file on GitHub · 228 lines

Changes

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.

  1. 5d ago First seen · 228 lines · 73 tokens per session scan A 6b66eb85e3a1

Subscribe to this mod's changes

qbcheck is a skill published in the GitHub repository 3awny/qship (2 stars, last pushed 2mo ago), licensed MIT. It adds 73 tokens to every session and 3,945 once invoked, about $0.0004 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.

Related

Other skills, from other repositories

remove-ai-marks

Strip multi-vendor AI provenance from owned files: hidden Unicode (Layer A), statistical sampling watermarks via rewrite (Layer B — always offer), and C2PA/EXIF/XMP/container metadata on PNG/JPEG/WebP/SVG/PDF/DOCX/ODT/HTML/MD. Covers Claude, Gemini/SynthID-class, OpenAI provenance surfaces, and open-LLM sampling…

ShadowAqueduct/watermark-remover · 135 tokens

clean-user-facing-text

Audit and finalize authorized natural-language text meant for readers: strip suspicious invisible Unicode, then rewrite prose while keeping facts, meaning, and the writer's voice. Use when the user asks to clean, humanize, polish, or finalize articles, manuscripts, reports, documentation, emails, product copy, UI…

ShadowAqueduct/watermark-remover · 121 tokens

oracle

Author an IMPL-BLIND spec-conformance oracle for an acceptance criterion the policy worklist (clad oracle --required) demands — an empty worklist means don't author unless the user explicitly asks. YOU spawn a blind sub-agent from a spec-only brief, then record it. Activate only when the connected project contains…

qwerfunch/cladding · 82 tokens

reviewer

Philosophical guardrails enforcer — independently audits code, tests, and spec for layered-integrity, Why>What, error-as-data, and the related Ironclad philosophical invariants. Activate only when the connected project contains spec.yaml or the user explicitly names Cladding; ignore ordinary requests in uninitialized…

qwerfunch/cladding · 68 tokens

changelog

Render release notes / a changelog from the cladding spec. Use when the user asks for release notes, a changelog, 릴리즈 노트, 변경 이력, or "what changed (since )" — run clad changelog --json for the deterministic shipped-changes manifest, then write the human-facing notes FROM it, sourcing every claim from a feature title or…

qwerfunch/cladding · 121 tokens

doctor

Diagnose Cladding runtime health — Claude Code hook liveness and version, CI package pinning, lifecycle governance, and sentinel-miss frequency by phase × cause × fallback. Use when hooks may be silent, CI may float across Cladding releases, scan or run results look thinner than expected, or before tuning the host…

qwerfunch/cladding · 93 tokens