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 dcramer/agents --skill testing-agentsgit clone --depth 1 https://github.com/dcramer/agentsWrote 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/dcramer/agents/testing-agents)<a href="https://agentmods.dev/skills/dcramer/agents/testing-agents"><img src="https://agentmods.dev/badge/skills/dcramer/agents/testing-agents/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/dcramer/agents/testing-agents"><img src="https://agentmods.dev/badge/skills/dcramer/agents/testing-agents.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00059 | $0.01145 |
| Opus 5.5 | $0.00024 | $0.00458 |
| Sonnet 5.5 | $0.00012 | $0.00229 |
| Haiku 4.5 | $0.00006 | $0.00114 |
Grade A, and why
testing-agents 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 — 57 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Testing Agents
Design the test from the contract outward. A trustworthy agent test is exact about fixed outcomes and flexible only where the product is genuinely variable.
Workflow
- Read the repository's test rules, commands, harnesses, and nearby coverage. State the user-visible or system contract and the regression the test should catch.
- Choose one primary owning layer:
- Use a unit test for pure fixed logic such as parsing, normalization, retry math, or scoring.
- Use a component or integration test for tool implementations, transport payloads, persistence, and fixed runtime lifecycle.
- Use an agent integration test or behavioral eval for model-dependent choices such as tool selection, routing, continuity, grounding, and reply strategy.
- Use an exact categorical eval when the model contract itself is a fixed enum such as
allow | ask | deny. Do not duplicate equal- or higher-fidelity coverage unless the new case owns a different contract or failure edge.
- Design one representative scenario per case. Use a plausible user input, one primary behavior, and observable pass or failure conditions. Name the trigger and expected outcome. Do not mirror every implementation branch or prescribe internal tools, commands, or reasoning in the input.
- Exercise production wiring. Run the production agent or application entry point with application-owned services. Use a live LLM when model behavior is under test. A repository-approved scripted model is appropriate only when the contract is fixed runtime behavior after a known response. Fake or replay only boundaries the repository permits, using realistic data.
- Classify every expected outcome before asserting it:
- Fixed invariant: exact enum, schema, required argument, row, request, artifact, or visible side effect. Assert it directly.
- Bounded structure: required subset or unordered effect where extra fields, steps, or order may vary. Use partial, set, or range matching.
- Semantic quality: meaning, completeness, tone, or grounding with multiple valid phrasings. Use a narrow judge only when no reliable direct oracle exists.
- Keep mixed cases mixed. If an agent must perform an exact mutation and then explain it naturally, directly assert the mutation and judge only the explanation. Use normalized harness projections for sessions, tool calls, messages, and artifacts instead of raw provider payloads or repo-local transcript schemas.
- Run the repository's focused command. Use its live-test credential behavior, bounded timeout, model configuration, and replay setup. Skip cleanly when credentials are absent only if that is repository policy. Report the exact command.
- Diagnose failures at the owning layer. Separate infrastructure failures from quality misses. Preserve useful run evidence. Do not weaken a deterministic oracle, broaden a rubric, or add retries before identifying whether the product, fixture, harness, judge, or threshold is wrong.
What ships with it
3 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 First seen · 57 lines · 59 tokens per session scan A f86a60a5d261
testing-agents is a skill published in the GitHub repository dcramer/agents (57 stars, last pushed today), licensed Apache-2.0. It adds 59 tokens to every session and 1,145 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 2026-09-03.
Other skills, from other repositories
research-engineer
An uncompromising Academic Research Engineer. Operates with absolute scientific rigor, objective criticism, and zero flair. Focuses on theoretical correctness, formal verification, and optimal implementation across any required technology.
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.
tika-eval-compare
Compare extracts from two Tika builds over a corpus to detect regressions in content, encoding, exceptions, and embedded-document handling. Use for "compare before/after extracts", "eval this change against the corpus".
cua-driver
Use Cua Driver for desktop or browser tasks that are awkward or unavailable through Bash/APIs, or when the user explicitly wants GUI interaction: app testing, visual bug reproduction, form filling, calendar entry, screenshots, and demo recording. Also covers Cua setup; not OpenAI Codex Computer Use or web research.
jetson-validate-image
Use after jetson-flash-image to run static BSP checks, on-target smoke/regression tests on a flashed DUT, or both. Not for build or flash steps. Triggers: validate bsp, on-target validation.
atmos-validation
Validate Atmos projects, components, arbitrary JSON Schema inputs, EditorConfig, and GitHub Actions; use affected-file selection and native CI annotations.