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 vinta/hal-9000 --skill refactor-skillgit clone --depth 1 https://github.com/vinta/hal-9000Wrote 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/vinta/hal-9000/refactor-skill)<a href="https://agentmods.dev/skills/vinta/hal-9000/refactor-skill"><img src="https://agentmods.dev/badge/skills/vinta/hal-9000/refactor-skill/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/vinta/hal-9000/refactor-skill"><img src="https://agentmods.dev/badge/skills/vinta/hal-9000/refactor-skill.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector warn
SkillSpector: 1 finding, up to medium
These are SkillSpector’s own severities. On a checked sample its high-severity flags on skills were ~96% false positives — a documented command, a public API, a “never do X” rule — so we show them as a caution to read, not a verdict. Why →
- medium Excessive Agency · line 45 Skill enables autonomous high-impact decisions without human-in-the-loop verification. Critical operations (destructive commands, financial transactions, data deletion) should require explicit user confirmation.Fix: Add human-in-the-loop confirmation for destructive, irreversible, or high-impact operations. Never auto-execute commands that modify files, send data, or alter system state.
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 | $0.00059 | $0.01454 |
| Opus 5.5 | $0.00024 | $0.00582 |
| Sonnet 5.5 | $0.00012 | $0.00291 |
| Haiku 4.5 | $0.00006 | $0.00145 |
Grade A, and why
refactor-skill 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 — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Overview
Refactor a skill. A skill spends two budgets. Its description is loaded into every session, so each word there must do triggering work. Its body enters the context on invocation and stays there for the rest of the session, so each line there must guide every path through the skill. Conditional material moves behind a pointer; unnecessary material leaves the skill.
Instructions
Use AskUserQuestion in Claude Code or request_user_input in Codex when available.
-
Pick the target. If the invocation names one, resolve it to its
SKILL.md. Otherwise ask which skill. For a full audit, readSKILL.mdand every sibling file in its directory. For a targeted change, readSKILL.md, affected files, and references needed to assess their interactions. That change sets the scope: the verdicts below cover the lines it adds or touches, the rest of the skill stays as it is, and anything noticed there is a follow-up. -
Fetch the guides. The platform and model guides below calibrate the delete, demote, and rewrite verdicts.
- Fetch one model guide, picked by the target skill's
model:frontmatter. Unset orinheritmeans the most capable model of the family running this skill:fableunder Claude,gpt-6-astraunder a GPT model. Read the matching reference and fetch the pages it lists:- Claude Code (
fable,opus,sonnet): references/claude-code.md. - Codex (
gpt-6-astra): references/codex.md.
- Claude Code (
- Fetch one model guide, picked by the target skill's
-
Audit. Give every frontmatter key, body instruction, and sibling file in scope exactly one verdict, judged against its bar below. Done when nothing in scope lacks a verdict.
- contradiction: conflicts with another instruction, or the frontmatter promises what the body does not deliver, such as a
context: forkskill whose body assumes conversation history. Record both sides. - delete: fails the no-op test, meaning the model the skill targets would behave the same without it, or the fetched pages say to remove it because it was written for an earlier model's habits. Judge defaults against those pages, not memory. Covers explanations of what the model already knows, emphasis that carries no instruction, and sibling files nothing points to.
- demote: real content that only some runs of the skill reach. Destinations: a file in the skill directory linked from
SKILL.md, one level deep; a script underscripts/when the steps are deterministic; a separate skill only when it needs its own trigger or another skill must reach it. Name the destination. - rewrite: right meaning, phrasing falls short of the fetched pages: a description that names features instead of triggers, a step with no completion criterion, a prohibition with no positive target, a rule shouted in capitals instead of explained. Also a rewrite when the line carries words that fail the no-op test on their own: a duplicate of another line, a definition of a term the model knows, or emphasis; a rationale or example stays, since both steer the model. Among wordings that pass the bar, the shortest wins. Voice follows the system prompt: the body addresses the model as "you" and names the human as "the user", so "I" or "my" for the human, and "Claude" or "the agent" for the model, are rewrites; words quoted as user speech keep their own voice.
- keep: earns its cost as written, and no shorter wording has the same effect.
- contradiction: conflicts with another instruction, or the frontmatter promises what the body does not deliver, such as a
What ships with it
3 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.
- today Changed a2fdcb0f415c
- 4d ago Changed a2198e1eb1c8
- 7d ago Changed · +2 lines a13f66804fe0
- 13d ago Changed a768203b4f7a
- 14d ago Changed · -7 tokens per session 6146c6b3b33d
- 19d ago Changed · -1 lines 9f0804fa89a7
- 22d ago Changed · +2 lines 50cbb123f56c
- 1mo ago Changed 32e8049a0e1f
- 1mo ago Changed · -7 lines aa7c7bbf1f6b
- 1mo ago First seen · 47 lines · 66 tokens per session scan A 82d951bda05e
refactor-skill is a skill published in the GitHub repository vinta/hal-9000 (139 stars, last pushed today), licensed MIT. It adds 59 tokens to every session and 1,454 once invoked, about $0.0002 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 2026-09-05.
Other skills, from other repositories
jarvis-setup
Use when helping someone install, configure, or debug a fresh clone of JARVIS (this repo) — especially "the mic doesn't work", "JARVIS says his language systems are down", any Firefox/browser question, login/auth failures, or Accessibility permission prompts. Carries facts about this project that were only learned by…
optimize-python-parallelism
Analyze and, when the task authorizes source optimization, refactor a concrete long-running Python entrypoint containing repeated independent work. Use for explicit Python speedup or parallelization requests and evidence-backed long-running entrypoints; skip ordinary Python edits, test execution alone, short scripts…
cirq
Google quantum computing framework. Use when targeting Google Quantum AI hardware, designing noise-aware circuits, or running quantum characterization experiments. Best for Google hardware, noise modeling, and low-level circuit design. For IBM hardware use qiskit; for quantum ML with autodiff use pennylane; for…
biopython
Comprehensive molecular biology toolkit. Use for sequence manipulation, file parsing (FASTA/GenBank/PDB), phylogenetics, and programmatic NCBI/PubMed access (Bio.Entrez). Best for batch processing, custom bioinformatics pipelines, BLAST automation. For quick lookups use gget; for multi-service integration use…
scikit-bio
Biological data toolkit. Sequence analysis, alignments, phylogenetic trees, diversity metrics (alpha/beta, UniFrac), ordination (PCoA), PERMANOVA, FASTA/Newick I/O, for microbiome analysis.
simpy
Build, inspect, test, and analyze bounded process-based discrete-event simulations with SimPy, including events, resources, interrupts, monitoring, replications, warm-up, and reproducible output analysis.