Borrowing it
Nothing to install: this file belongs to NVlabs/ASPIRE. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/NVlabs/ASPIRE/main/.claude/skills/iterative-debugging/behavior/SKILL.mdgit clone --depth 1 https://github.com/NVlabs/ASPIREWrote 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/nvlabs/aspire/behavior)<a href="https://agentmods.dev/skills/nvlabs/aspire/behavior"><img src="https://agentmods.dev/badge/skills/nvlabs/aspire/behavior/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/nvlabs/aspire/behavior"><img src="https://agentmods.dev/badge/skills/nvlabs/aspire/behavior.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 | $0.00099 | $0.01907 |
| Opus 5.5 | $0.00040 | $0.00763 |
| Sonnet 5.5 | $0.00020 | $0.00381 |
| Haiku 4.5 | $0.00010 | $0.00191 |
Grade A, and why
iterative-debugging-behavior 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 21d 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 — 123 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Block-by-Block Fix Loop on BEHAVIOR-1K
The parent skill (../SKILL.md) writes one whole program per task, freezes it, and replays that frozen program across the held-out seeds. That protocol does not apply here. BEHAVIOR tasks are long-horizon: the robot starts in another room, cannot see the target, and must search, drive, look down and grasp. A program authored up front is guessing at every stage past the first.
So on this benchmark the policy is grown, not authored:
- Write one
# Code block Nsection. - Run the whole
policy.pyon the seed, from a fresh reset. - Read what actually happened: where the base ended up, what was localized, what the hand holds.
- Write the next block against that observed state, and run the whole program again.
- Repeat until the seed succeeds or the budget is spent.
The simulator resets on every attempt. Only policy.py persists and accumulates blocks.
What carries across seeds is the skill library, not any policy. Each seed's policy.py is
disposable. Stage 2 measures whether the frozen library is enough for a fresh agent to solve an
unseen seed, not whether one program generalizes.
The library is shared by both tasks. It lives at $RUN_ROOT/behavior/skill-library/, above the
suite directories, and a task's Stage 1 extends what the previous task left. This is a deliberate
departure from ASPIRE, which keeps one library per task. The reason is that most of what a campaign
learns here is about the harness, the robot and the simulator rather than about one object, and
re-discovering it per task wastes a bootstrap seed each time. It also asks a more interesting
question: does a library built on one task help on a different one.
It leaks nothing, because the partition is per task. Knowledge measured on the radio's seeds says nothing about a soda-can instance, so the can's evaluation seeds stay unseen.
The cost is that every entry must carry the task it was measured on, and an entry from another task is a hypothesis rather than a rule until this task's seeds confirm it.
| Difference from the parent | Consequence |
|---|---|
| Long horizon, target not visible at reset | Blocks are written against observed state, never predicted state |
| The deliverable is a library | Stage 1 promotes skills after every seed; policies are disposable |
| Stage 2 is agentic | A fresh agent per evaluation seed writes its own blocks against the frozen library |
Frames: odom is fixed where the base stood at reset |
Localized objects stay valid after the base drives |
| Episodes cost minutes, not seconds | Every attempt replays every earlier block; keep blocks lean |
| Isaac Sim, not MuJoCo | OMNIGIBSON_* variables, one Isaac process per GPU, exit 139 after a complete run is normal, judge by outcome.json |
| Planning runs in-process on cuRobo | MotionStrategy() already means cuRobo; PyRoKi and port 8118 are not involved |
| Two arms, a torso and a base | behavior.* extensions move base and torso; RobotAction moves arms |
What ships with it
10 files 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.
- 21d ago First seen · 123 lines · 99 tokens per session scan A 19d2d3b78755
iterative-debugging-behavior is a skill published in the GitHub repository NVlabs/ASPIRE (959 stars, last pushed 22d ago), licensed Apache-2.0. It adds 99 tokens to every session and 1,907 once invoked, about $0.0004 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 2026-10-17.
iterative-debugging-behavior compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| iterative-debugging-behaviorNVlabs/ASPIRE | Skill | 959 | 22d ago | Apache-2.0 | A | 99 | Claude Code |
| traceYeachan-Heo/oh-my-claudecode | Skill | 41k | yesterday | MIT | A | 18 | Claude Code |
| pos-tagging-fixesspencermountain/compromise | Skill | 12k | 2d ago | MIT | A | 72 | Claude Code, Codex |
| smiles-validationsynthetic-sciences/openscience | Skill | 4.3k | 2d ago | Apache-2.0 | A | 24 | Claude Code, Codex |
| nonlinear-solversbeita6969/ScienceClaw | Skill | 1.0k | yesterday | MIT | A | 63 | Claude Code |
| numerical-stabilitybeita6969/ScienceClaw | Skill | 1.0k | yesterday | MIT | A | 61 | Claude Code |
| linear-solversbeita6969/ScienceClaw | Skill | 1.0k | yesterday | MIT | A | 55 | Claude Code |