SkillsBench is a benchmark for measuring how effectively AI agents use modular skills—folders containing instructions, scripts, and resources—to complete specialized tasks. It helps researchers and developers evaluate both skill quality and agent behavior, including tasks that require combining multiple skills. The catalogue’s skills and instructions are evaluated as part of this workflow.
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 benchflow-ai/skillsbench --skill stepinfo-3dgit clone --depth 1 https://github.com/benchflow-ai/skillsbenchWrote 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/benchflow-ai/skillsbench/stepinfo-3d)<a href="https://agentmods.dev/skills/benchflow-ai/skillsbench/stepinfo-3d"><img src="https://agentmods.dev/badge/skills/benchflow-ai/skillsbench/stepinfo-3d/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/benchflow-ai/skillsbench/stepinfo-3d"><img src="https://agentmods.dev/badge/skills/benchflow-ai/skillsbench/stepinfo-3d.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 | $0.00068 | $0.00824 |
| Opus 5 | $0.00034 | $0.00412 |
| Sonnet 5 | $0.00014 | $0.00165 |
| Haiku 4.5 | $0.00007 | $0.00082 |
Grade A, and why
stepinfo-3d 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 8d 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 — 58 lines — stays where its author put it; the contents beside it link to each section on GitHub.
3D Step-Response Metrics (stepinfo_3d)
When to Use
| Scenario | Metric to use |
|---|---|
| Pure z-step (hover, takeoff, land) | 1D stepinfo on z signal |
| Diagonal flight (x, y, z all change) | stepinfo_3d on 3D Euclidean distance |
| Circular / figure-eight trajectory | Neither — use RMS error or cumulative error |
1D metrics break for diagonal flight because the axes are coupled — thrust that corrects x also affects y and z.
Metrics Defined
| Metric | Definition |
|---|---|
| Rise time | First time 3D distance to target ≤ 10% of initial distance |
| Settling time | Last time 3D distance exceeds settling_threshold × initial_distance |
| Overshoot % | Max distance from target after first entering the settling band, as % of initial distance |
| Steady-state error | Final 3D Euclidean distance from target [metres] |
Implementation Logic
Given pos_actual (3, n), pos_target (3,), and time vector t (n,):
- Compute
dist[k] = ||pos_actual[:, k] − pos_target||₂for each timestep. - If
dist[0] < 1e-6(already at target), return all zeros. - Rise time: scan forward and record the first
t[k]wheredist[k] ≤ 0.1 * dist[0]. - Settling time: scan backward and record the last
t[k]wheredist[k] > settling_threshold * dist[0](default threshold = 0.02). - Overshoot: after the drone first enters the settling band, track the maximum
dist[k]seen. Express asmax_post_entry / dist[0] * 100. If the settling band is never entered, return 0. - Steady-state error:
dist[-1].
Return a dict with keys RiseTime, SettlingTime, Overshoot_pct, SteadyStateError.
Usage in Simulation
from stepinfo_3d import stepinfo_3d
pos_final_desired = waypoints[0:3, -1] # last waypoint
metrics = stepinfo_3d(actual_state_matrix[0:3, :], pos_final_desired, time_vec)
for k, v in metrics.items():
print(f' {k}: {v:.4f}' if isinstance(v, float) else f' {k}: {v}')
Limitations
- Assumes point-to-point flight — the drone starts away from a fixed target and converges. For circular trajectories, use RMS or cumulative error instead.
dist_initialis the distance att[0]. If the drone starts at the target (hover command), all metrics return0.- Overshoot is defined by distance, not by crossing the target in one axis — the drone must physically move farther from the target after settling to register overshoot.
- If the settling band is never entered (common for very short commands where d0 is small, making
band = 0.02 × d0only a few centimetres),Overshoot_pctreturns0.0— the drone approached the target without oscillating past it.
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.
- 8d ago First seen · 58 lines · 68 tokens per session scan A e139d7322aaa
stepinfo-3d is a skill published in the GitHub repository benchflow-ai/skillsbench (1,764 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 68 tokens to every session and 824 once invoked, about $0.0003 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 2026-09-03.
Other skills, from other repositories
exploded-view-illustration
Techniques for drawing isometric or orthographic exploded-view hardware diagrams programmatically using Pillow, with annotation leader lines and layer separation.
membrowse-integrate
Integrate MemBrowse memory tracking into a project that produces ELF binaries. Use when the user wants to set up MemBrowse, add memory analysis GitHub workflows, create membrowse-targets.json for tracking memory usage, or add a MemBrowse badge to the README. Works with embedded firmware (STM32, ESP32, nRF, RISC-V) and…
pcbway
PCBWay PCB fabrication and assembly — turnkey/consigned assembly, design rules, ordering workflow. Alternative to JLCPCB for manufacturing. Use with KiCad. Use this skill when the user mentions PCBWay, needs turnkey assembly (PCBWay sources parts by MPN), has parts not available on LCSC, needs assembled boards with…
unifi-protect
How to manage UniFi Protect cameras and NVR — view cameras, smart detections, Find Anything detection search, recordings, snapshots, lights, sensors, Known Faces, license plates, and the Alarm Manager. Use this skill when the user mentions UniFi cameras, security cameras, NVR, recordings, motion detection, person…
akg-agents
An agent workflow for developing, generating, testing, and tuning AKG operators. AKG is a toolkit for generating optimized computational kernels.
fpga
FPGA development guidelines covering Vivado, SystemVerilog, timing closure, AXI interfaces, and hardware optimization.