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 mehdiforoozandeh/crux --skill crux-autopilotgit clone --depth 1 https://github.com/mehdiforoozandeh/cruxWrote 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/mehdiforoozandeh/crux/crux-autopilot)<a href="https://agentmods.dev/skills/mehdiforoozandeh/crux/crux-autopilot"><img src="https://agentmods.dev/badge/skills/mehdiforoozandeh/crux/crux-autopilot/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/mehdiforoozandeh/crux/crux-autopilot"><img src="https://agentmods.dev/badge/skills/mehdiforoozandeh/crux/crux-autopilot.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 | $0.00272 | $0.03357 |
| Opus 5.5 | $0.00109 | $0.01343 |
| Sonnet 5.5 | $0.00054 | $0.00671 |
| Haiku 4.5 | $0.00027 | $0.00336 |
Grade A, and why
crux-autopilot 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 — 208 lines — stays where its author put it; the contents beside it link to each section on GitHub.
crux-autopilot — write the flight plan, then stop
A flight plan carries twenty slots plus a baseline, and crux auto run refuses a plan short
of any of them. This skill gets the PI from "I want to search this" to a plan that passes
crux auto check. It asks for four of those slots. Everything else is derived from an
answer already given, measured by a command, or defaulted and stated back in one line.
The six acts — four asks, two approvals
| # | the skill's turn | why it cannot be skipped |
|---|---|---|
| 1 | ask — the anchor question, and the goal in one falsifiable sentence | the frame is the PI's; nothing can derive it |
| 2 | approve — the null, proposed by crux-null |
a ruling, not a value; spec 05 requires the PI's yes |
| — | derived — the checks, their kinds, the combination rule, the control | crux-verifiables, against the approved null |
| — | derived — the objective address | read off the discriminating check just written |
| 3 | ask — which direction is better, and the bar | direction is the one thing that must never be guessed |
| 4 | ask — the harness, in one turn: scorer command, run command, frozen paths, writable paths, budget ceiling, and the workers' model and effort |
only the PI knows their repo and their budget; model and effort set how long every attempt takes, so they are never defaulted — the engine refuses a plan without them |
| 5 | ask — the PI's standing guidance: their priors, the analogies they see, the routes they rate, the dead ends | the one input the search cannot find for itself |
| — | measured — the baseline, and the published rivals | the scorer is run; the literature is searched |
| — | defaulted — the remaining slots | stated back in one line, not asked |
| 6 | approve — crux-design's verdict, then the PI's own crux auto approve |
the PI's signature |
The ordering that is load-bearing
The objective address is read off a check, never chosen: the bar-and-direction turn comes
after crux-verifiables has written the checks, so the address comes out of a check that
already exists and the engine's discriminates lint passes as a consequence rather than as
a hurdle. A PI who picks the number first and writes a check around it afterwards produces
a plan that passes every lint and is exactly the failure the spec forbids: the search clears
a bar the hypothesis was always going to clear, thousands of times, at cost, and teaches
nothing.
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 d4c1b606f24f
- 1mo ago Changed · +95 lines · +38 tokens per session 5a0bf1f85100
- 1mo ago First seen · 113 lines · 234 tokens per session scan A dc0b88fbe4d1
crux-autopilot is a skill published in the GitHub repository mehdiforoozandeh/crux (18 stars, last pushed today), licensed MIT. It adds 272 tokens to every session and 3,357 once invoked, about $0.0011 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 2026-09-21.
Other skills, from other repositories
academic-writing
You must use this when producing any research prose — literature reviews, syntheses, analyses, methodology descriptions, discussion sections, abstracts, or any written output intended for an academic audience.
systematic-review
You must use this when conducting PRISMA-standard systematic reviews, protocol development, or Risk of Bias assessment.
hypothesis-testing
You must use this when formulating testable hypotheses, designing experimental controls, or defining falsification criteria.
quantitative-analysis
You must use this when selecting statistical tests, interpreting effect sizes, or conducting power analysis.
research-synthesis
You must use this when merging findings from multiple studies into a coherent narrative with grounded evidence.
literature-review
You must use this when synthesizing existing knowledge, identifying research gaps, or tracing the evolution of scientific ideas.