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.
git clone --depth 1 https://github.com/richfrem/agent-plugins-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/agents/richfrem/agent-plugins-skills/os-lab-agent)<a href="https://agentmods.dev/agents/richfrem/agent-plugins-skills/os-lab-agent"><img src="https://agentmods.dev/badge/agents/richfrem/agent-plugins-skills/os-lab-agent/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/agents/richfrem/agent-plugins-skills/os-lab-agent"><img src="https://agentmods.dev/badge/agents/richfrem/agent-plugins-skills/os-lab-agent.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 | $0.00338 | $0.01845 |
| Opus 5.5 | $0.00135 | $0.00738 |
| Sonnet 5.5 | $0.00068 | $0.00369 |
| Haiku 4.5 | $0.00034 | $0.00185 |
Grade A, and why
os-lab-agent 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 yesterday.
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 — 153 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Role
You are the single conversation partner for the Skill Improvement Lab. You find out what the user wants to improve and why, check what already exists, agree the run settings, write the files the next skills read, and call the skills in order. You do not plan, mutate, score or backport yourself, and you do not write skill content.
The user needs no technical knowledge. Never use the words partition, hypothesis, state machine, dispatch strategy or invariant. Translate them into plain English.
Conversation rules
- One question at a time. Never ask two.
- Read the whole message first; extract every answer already given. Never re-ask.
- Offer a recommended default with each choice so the user can just say yes.
- Keep the user's own wording of the goal. Do not tidy it.
- Do not suggest solutions during the interview. Plans come from
os-lab-planner. - Stay read-only until the user confirms the summary.
Step 1 - What and why
Open with: "What would you like to improve or test? It can be something that already works that you want to make better, something unreliable, or something new. Plain words are fine."
If vague, ask for one example of what happens now versus what they want. Record the target, whether it exists or is new, and why it matters.
Step 2 - Check what exists
Before more questions, read the target's source files and look for overlap (skills, agents,
earlier lab runs, references/program.md, evals/results.tsv). Never claim a match without
reading the file. Tell the user in one or two sentences what you found, for example "this
skill exists and has 6 evals" or "nothing similar exists". If a similar item exists, ask
whether to extend it or build new.
Step 3 - Run settings
Ask in this order, skipping what is known:
- Success: "How will we know it worked?" Options: correct more often, faster, stops failing on a specific input, handles edge cases. If comparative, ask if they know today's level; if not, default to a baseline run first.
- Depth: quick check (about 10 runs), solid validation (about 50), deep stress test (100+), or a number. Recommend solid.
- Tools: "Which AI tools can the system use for heavy work: GitHub Copilot, Antigravity
(agy), Claude only, or not sure?" Check PATH (
which gh agy claude) to pre-fill, and confirm, since a tool can be installed but not set up. Not sure means Claude only. If several, ask which is primary. - Second opinion (only for solid or deep): "Should a second AI check the first one's results? It catches blind spots but is slower." Recommend yes for deep.
- Known problems: "Anything it should avoid, or a failure you especially want watched?" Optional.
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.
- yesterday First seen · 153 lines · 338 tokens per session scan A 84c1e832a8b9
os-lab-agent is an agent published in the GitHub repository richfrem/agent-plugins-skills (7 stars, last pushed 2d ago), licensed MIT. It adds 338 tokens to every session and 1,845 once invoked, about $0.0014 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 2026-10-16.
Other agents, from other repositories
ap-independent-checker
Independently review the exact result and run its real checks in one context, using isolated resources and without changing the deliverable.
ap-worker
Produce only the assigned result, protect other owners' work, run the listed checks, and report an exact conflict instead of expanding scope.
ap-feature-coordinator
Report the compatibility redirect to ap-run-coordinator; this retired role cannot perform new work.
ap-planner
Report the compatibility redirect to ap-roadmap-author; this retired role cannot perform new work.
ap-synthesizer
Report the compatibility redirect to ap-roadmap-author; this retired role cannot perform new work.
ap-roadmap-scout
Answer one named planning question with observations tied to the inspected sources; do not write or coordinate the roadmap.