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 mdrideout/junjo --skill junjo-evaluationgit clone --depth 1 https://github.com/mdrideout/junjoWrote 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/mdrideout/junjo/junjo-evaluation)<a href="https://agentmods.dev/skills/mdrideout/junjo/junjo-evaluation"><img src="https://agentmods.dev/badge/skills/mdrideout/junjo/junjo-evaluation/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/mdrideout/junjo/junjo-evaluation"><img src="https://agentmods.dev/badge/skills/mdrideout/junjo/junjo-evaluation.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 | $0.00083 | $0.02551 |
| Opus 5.5 | $0.00033 | $0.01020 |
| Sonnet 5.5 | $0.00017 | $0.00510 |
| Haiku 4.5 | $0.00008 | $0.00255 |
Grade A, and why
junjo-evaluation 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 — 259 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Junjo Evaluation
Operate Junjo Evaluation from the application repository using the installed
junjo SDK and CLI. Do not require access to the Junjo source repository.
Operating contract
Treat a product-quality objective such as “evaluate local-place realism” as sufficient direction to operate the complete baseline workflow. Own target and evaluator discovery, scenario design, temporary JSON artifacts, Studio dataset operations, run execution, identifier tracking, evidence retrieval, analysis, and the final report.
Do not ask the developer to write JSON, choose CLI flags, copy identifiers, poll evidence, or run commands. Ask only when product intent is materially ambiguous or a real prerequisite or authorization is missing.
For a baseline request, measure current committed behavior and do not modify application behavior. Modify prompts or code, run tests, or create commits only when the developer authorizes an improvement iteration. Never commit unrelated work.
Boundaries
- Use
junjo eval explainas the installed command, configuration, and output reference. Use the current command--helpfor exact syntax. Do not reproduce CLI mechanics with ad hoc HTTP or shell clients. - Use
junjo.evaluationandjunjo.studiowhen application code or direct Python integration is required. Do not copy Studio clients, DTOs, runners, or generic evaluation mechanics into the application. - Keep
JUNJO_AI_STUDIO_CLI_TOKENseparate fromJUNJO_AI_STUDIO_API_KEY. Never print, persist, or pass either secret as a routine command argument. - Preserve the application's real OpenTelemetry service identity. Evaluation classification augments ordinary application telemetry; it does not create a fake evaluation service.
- Use bounded SDK/CLI evidence and comparison queries rather than Studio's raw observability routes.
- Let each Junjo executor process its Attempts sequentially. Parallel experiments use independent Runs with distinct request keys and one runner owning each Run; they do not share a Run between concurrent executors.
What ships with it
1 file 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 · +3 lines 825f54522f7f
- 1mo ago Changed · +21 lines 2d802354f8c5
- 1mo ago Changed · +34 lines dfc5ce6e41a8
- 1mo ago First seen · 201 lines · 83 tokens per session scan A d54cee925029
junjo-evaluation is a skill published in the GitHub repository mdrideout/junjo (20 stars, last pushed 17d ago), licensed Apache-2.0. It adds 83 tokens to every session and 2,551 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.
Other skills, from other repositories
credit-note-fixer
Fix the tiny credit-note formatting bug and rerun the exact targeted test command.
verify-omnigent
Drive Omnigent the way a user does and prove a behavior with recorded evidence, using an isolated server, runner, and mock model plus a feature map of every user entry point. Load before reproducing a user-facing bug, before claiming a fix works, or when reviewing whether a change covered every surface (session vs.…
implement-universal
Harness-agnostic version of /implement. Drives a single workshop ticket through the SWE→Tester loop in ONE conversation, with the role prompts bundled as agents/software-engineer.md and agents/tester.md instead of being launched as subagents. Resolves the ticket from implementyourself/tasks/, creates an…
playwright-cli
Automates browser interactions for web testing, form filling, screenshots, and data extraction. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from web pages.
resolve-author-fix
Find the root cause, implement a focused fix, and prove behavior with targeted tests and recordings.
resolve-repro-audit
Audit a recovered repro patch, assertions, and behavioral baseline before authoring or reviewing.