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 TimothyHan/qa-buddy-skills --skill evalgit clone --depth 1 https://github.com/TimothyHan/qa-buddy-skillsWrote 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/timothyhan/qa-buddy-skills/eval)<a href="https://agentmods.dev/skills/timothyhan/qa-buddy-skills/eval"><img src="https://agentmods.dev/badge/skills/timothyhan/qa-buddy-skills/eval/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/timothyhan/qa-buddy-skills/eval"><img src="https://agentmods.dev/badge/skills/timothyhan/qa-buddy-skills/eval.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 2 findings, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium MCP Rug Pull · line 32 npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.Fix: Pin the version: npx @scope/[email protected]
- medium MCP Rug Pull · line 119 npx commands without a version suffix (e.g. @1.0.0) create a rug-pull risk if the upstream server is compromised and publishes a malicious update.Fix: Pin the version: npx @scope/[email protected]
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 | $0.00130 | $0.02245 |
| Opus 5.5 | $0.00052 | $0.00898 |
| Sonnet 5.5 | $0.00026 | $0.00449 |
| Haiku 4.5 | $0.00013 | $0.00225 |
Grade A, and why
eval 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 1mo 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 — 197 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/qa-eval: Skill Eval Testing
You run eval fixtures against a skill to verify it produces correct output. Each fixture declares its mode:
simulate(default) — read the skill's instructions, simulate the scenario, check assertions against the hypothetical output. For skills whose output is prose (reports, verdicts, plans).execute— actually run the skill against the local fixture app, then grade the artifacts it produced by executing them (npx playwright test, greps, file checks). For skills whose output is code (/qa-e2e-setup,/qa-e2e-pom,/qa-e2e-write). A generated artifact passes only by running, never by looking right.--rubric(RFC 0005) —/qa-eval <skill> --rubricdelegates tonode bin/eval.js run <skill>: the skill runs headless on the target model, a separate Opus judge scores each artifact againsttests/rubric.json, and controls are judged first. The summary is the generatedreport.md(per-criterion means, spread, floor breaches, verdict) — never re-grade its scores by hand. For a skill whose rubric is calibrated, simulate-mode fixtures are deprecated as a quality signal; they stay as the free structural gate.
Constraints
- One skill at a time. Don't run all skills in one session — context window.
- Follow the skill's instructions exactly. When simulating or executing, you are that skill. Apply its constraints, phases, and methodology references.
- Grade honestly. If an assertion fails, report it — don't rationalize a pass.
- Assertions are literal.
contains "READY"means the string "READY" appears in the output. Don't interpret loosely. - Simulate mode never touches the real environment — no browsers, no Jira, no file writes.
- Execute mode touches ONLY the local fixture app and a scratch workspace. Never external systems: no Jira, no real staging URLs, no network beyond
localhost. All generated artifacts go in a throwaway workspace directory, never the QABuddy repo. - Never open
ANSWER-KEY.mdwhile acting as the skill under eval. The answer key is for grading only. Discovery must happen against the running app. Reading the key mid-simulation invalidates the fixture — report it as a harness error, not a pass. - Scripted user responses replace the user. Execute fixtures include
user_responsesfor the skill's interactive gates. Answer exactly as scripted. If the skill never asks at a gate where a response was scripted, that's a finding — several fixtures assert the skill does pause.
What ships with it
10 files 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.
- 1mo ago Changed · +10 lines 6badafcb1de3
- 1mo ago First seen · 187 lines · 130 tokens per session scan A 1202194cfff0
eval is a skill published in the GitHub repository TimothyHan/qa-buddy-skills (9 stars, last pushed 19d ago), licensed Apache-2.0. It adds 130 tokens to every session and 2,245 once invoked, about $0.0005 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
loop-engineering
A guide to building reliable loops for AI agents that plan, act, check their work, and continue until a task is truly complete. It includes independent verification and a proposer-reviewer pattern in which one agent's work is checked using outside evidence.
agent-evaluation
A guide to measuring whether an AI agent works reliably on the tasks that matter. It covers repeatable test environments, program checks, human-like scoring rules, failure analysis, and experiments that separate model problems from system problems.
bad-case-to-dpo
A guide to turning failed production agent runs into training and evaluation data. It explains how to find the first mistake, create preference examples for DPO (Direct Preference Optimization), and check that a fix does not damage other tasks.
eval-dataset-design
A guide to creating reliable test sets for AI agents. Each evaluation task includes the request, the simulated user's behaviour, a reset starting state, and a success check that can be independently verified.
qawolf-cli
Manage QA Wolf through the qawolf CLI. Use when asked to create, update, or list coverage requests, bug reports, or maintenance reports; start a run of flows or tags on the QA Wolf platform or read a run's results; list, set, or delete environment variables; manage environments, flows, or tags; run or list flows…
test-review
A review checklist for newly written or changed automated tests using TypeScript and Playwright. Playwright is a tool for testing web browsers, and automated tests check software without repeating the steps by hand.