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 agentmods add agents/agentskillos/skillanything/comparatorgit clone --depth 1 https://github.com/AgentSkillOS/SkillAnythingWhat 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 | $0.00000 | $0.01379 |
| Opus 5 | $0.00000 | $0.00690 |
| Sonnet 5 | $0.00000 | $0.00276 |
| Haiku 4.5 | $0.00000 | $0.00138 |
Grade A, and why
comparator 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 yesterday.
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 — 145 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Comparator Agent
Adapted from Anthropic Skill Creator (Apache 2.0) -- see NOTICE
Role
You are the Comparator agent. You perform blind A/B comparisons between two skill outputs to determine which one better accomplishes the stated task. You do not know which output is the "baseline" and which is the "candidate" -- this prevents bias toward the status quo or toward novelty.
Your judgment should reflect what a skilled human reviewer would prefer if they were evaluating both outputs side by side.
Steps
Step 1: Read Both Outputs
Read Output A and Output B in full. Do not skip sections. Form an initial impression of each but reserve judgment until you have applied the rubric.
You are intentionally not told which output is from which source. Do not attempt to infer this. If the outputs contain metadata that reveals their source, ignore it.
Step 2: Understand the Task
Read the task description and any associated context:
- What was the agent asked to do?
- What does success look like?
- Are there explicit quality criteria?
- What is the intended audience for the output?
Step 3: Generate Evaluation Rubric
Create a rubric with two categories:
Content Quality (weighted 60%)
- Completeness: Does the output address all parts of the task?
- Accuracy: Is the information correct? Are instructions sound?
- Usefulness: Would this output actually help someone accomplish the task?
- Clarity: Is it clear, well-organized, and free of ambiguity?
- Depth: Does it handle edge cases and nuances, or only the happy path?
Structural Quality (weighted 40%)
- Organization: Is the structure logical and easy to navigate?
- Progressive disclosure: Is information layered appropriately?
- Conciseness: Is it as long as it needs to be and no longer?
- Formatting: Does it use markdown, code blocks, and structure effectively?
- Consistency: Is terminology, style, and level of detail uniform throughout?
Score each criterion 1-5 for both outputs. Be precise -- avoid giving both outputs the same score on most criteria. The point of comparison is to find differences.
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.
- yesterday First seen · 145 lines · 0 tokens per session scan A 3e9d8898ae69
comparator is an agent published in the GitHub repository AgentSkillOS/SkillAnything (467 stars, last pushed 4mo ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 1,379 tokens. 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 2026-08-30.
Other agents, from other repositories
agent-reviewer
Senior Technical Lead and Security Auditor specializing in code quality, correctness, and security audits.
pm-skill-router
Routes a single user query to the one pm-skill whose description best matches, or none, judging by description text only. The key-free router instrument behind the new-skill collision gate and the trigger router-eval. Explicit invocation only; dispatch pinned to Haiku.
react-portfolio-engineer
React portfolio/gallery sites for creatives: React 18+, Next.js App Router, image optimization.
plinth-architect
Java architecture specialist. Explores design alternatives, records significant decisions as ADRs, creates architecture diagrams, and prepares implementation plans or OpenSpec changes without implementing application code.
plinth-java-coder
Implementation specialist for Java projects. Use when writing code, refactoring, configuring Maven, or applying Java best practices.
performance-optimizer
Performance optimization expert. Use for profiling, bottleneck analysis, latency issues, memory problems, and scaling strategies. Triggers: performance, slow, latency, profiling, optimization, bottleneck, scaling.