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 donald-ada/workinggenius --skill wondergit clone --depth 1 https://github.com/donald-ada/workinggeniusWrote 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/donald-ada/workinggenius/wonder)<a href="https://agentmods.dev/skills/donald-ada/workinggenius/wonder"><img src="https://agentmods.dev/badge/skills/donald-ada/workinggenius/wonder.svg" alt="Measured on agentmods" 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 | $0.00072 | $0.00814 |
| Opus 5 | $0.00036 | $0.00407 |
| Sonnet 5 | $0.00014 | $0.00163 |
| Haiku 4.5 | $0.00007 | $0.00081 |
Grade A, and why
wonder 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 today.
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 — 29 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Wonder
The genius of questioning. Most failed work fails here first: the agent built exactly what was asked, and what was asked was not what was wanted.
The concept: confirm the problem before designing anything. Interview the user — live dialogue, always; a model interviewing itself confirms nothing. How you interview is your judgment:
- Ask only what the user alone can answer — everything the repo, the history or the docs can settle is homework, not a question. Attach your best recommendation to every question, and say the price when an answer forks the cost. The one question no homework can settle is how they will judge it done — the run they would make, the number they would look at, the case they would try first — and it goes into the problem statement in their words, because a success the interviewer wrote and they nodded at confirms the interviewer's picture, and the bar they actually hold arrives otherwise at close-out, when meeting it costs the most.
- Ask in rounds — a batch of independent questions in one message, the next round grown from the answers. The one-question drip costs round-trips and patience; a wall of everything is a form, not an interview.
- Telling them the story you'd build and letting them correct it is usually cheaper for them than answering from zero.
- Ask the question behind the request — a feature ask is often a solution wearing a problem's clothes. "Don't build this" is a successful outcome.
- The interview ends when nothing load-bearing is still assumed — not when your questions run out, and not when the user nods at a statement vague enough to nod at. Ending early is the user's line to say ("enough — go with your recommendations"), never an exit you offer them; when they say it, record what you assumed.
- An architecture question surfacing mid-interview is a fork, not small talk — answered casually, it becomes a defaulted decision. Name
/architectand let the user call it; the interview recommends, it never dispatches.
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.
- today Changed ffd670beccbd
- 2d ago Changed · +8 lines 207053a90dbd
- 7d ago First seen · 21 lines · 72 tokens per session scan A 42f220cd1c96
wonder is a skill published in the GitHub repository donald-ada/workinggenius (6 stars, last pushed yesterday), licensed MIT. It adds 72 tokens to every session and 814 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.
Other skills, from other repositories
self-improve
Self-improve this skills repo — surface and (safely) apply quality improvements across four parallel read-only lenses (Clarity, Leanness, Coverage, Ergonomics). Sibling of /self-audit — audit fixes rule violations, improve raises quality where no rule is broken. Local meta-skill, independent of the /task: pipeline.
self-audit
Self-audit this skills repo against CLAUDE.md invariants, the artifact contract, and README/CLAUDE.md/docs sync via three parallel read-only subagents. Local meta-skill — independent of the /task: pipeline.
grill
Interrogate a plan or decision one question at a time before capture, keeping a decision-plus-rationale ledger, then route to the right capture skill.
dw-doctor
Read-only environment diagnostic for a solo-lane repo: whether the tools the hooks and skills assume are installed, and whether the guardrails and the pre-commit really fire here or are silently skipping, each gap reported with a copy-paste fix.
dw-handoff
Compact the live session into .ai/work/ /HANDOFF.md — how far into the current task you are, what is applied but uncommitted, and which dead ends are already ruled out, so the next context window resumes mid-task. Explicit-invoke only.
dw-next
Continue the change you are in the middle of, from where the last session had to leave it: catch up on where it stands and what is left, then build every unticked task, one commit each. Its state is read back from the doc, never from the conversation, so it survives a /clear.