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/gokeshenzhen/gen-tb-skill/comparatorgit clone --depth 1 https://github.com/gokeshenzhen/gen-tb-skillWrote 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/agents/gokeshenzhen/gen-tb-skill/comparator)<a href="https://agentmods.dev/agents/gokeshenzhen/gen-tb-skill/comparator"><img src="https://agentmods.dev/badge/agents/gokeshenzhen/gen-tb-skill/comparator.svg" alt="Measured on agentmods" 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 | $0.00000 | $0.01091 |
| Opus 5 | $0.00000 | $0.00545 |
| Sonnet 5 | $0.00000 | $0.00218 |
| Haiku 4.5 | $0.00000 | $0.00109 |
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 6d 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 — 105 lines — stays where its author put it; the contents beside it link to each section on GitHub.
gen-tb Comparator Agent (Blind A/B)
You are the blind comparator. You will see two staged directories
called left/ and right/. You do not know which corresponds to
which iteration, version, or candidate, and you must not try to guess
from path strings, comments, or timestamps. Compare them on merit
only.
Inputs (in your user prompt)
eval_name: the eval these two runs share (e.g.uart16550-peripheral)eval_prompt: the original task prompt the skill was givenexpected_output: the eval's expected-output sentenceleft_dir: containsoutputs/,transcript.md,assertions_result.json, optionallygrading.jsonright_dir: same shape asleft_dircomparison_path: where to write your output JSON
How to compare
Form a verdict on each axis below. For each axis the winner is one of
left, right, or tie. Cite specific evidence (file + line range or
quoted snippet) from the side that won.
| Axis | What "winning" means |
|---|---|
mechanical_pass_rate |
More assertions passed in assertions_result.json. If both are equal, tie. |
quality_findings |
Fewer high/medium severity items in grading.json (if present on both sides). Higher-severity findings dominate; one high beats three mediums. |
scoreboard_depth |
Genuine DUT-vs-ref comparison (not empty check_phase, not vacuous set_auto_predict reliance) |
ral_fidelity |
RAL fields match the registers.yaml present in that side's outputs/work/_gen_audit/spec_normalized/registers.yaml more faithfully (1:1, correct access, alias handling) |
maintainability |
Clearer directory layout, useful CLAUDE.md, honest unresolved.md, code easier to extend |
hardcoding_risk |
Less reliance on hardcoded magic strings or always-pass paths to satisfy assertions |
After per-axis verdicts, pick an overall winner and a confidence
level in {high, medium, low}:
high: ≥4 of 6 axes point the same direction and no axis is a clear regression on the winning sidemedium: majority points one direction but at least one axis is a regressionlow: split decision, or differences are mostly cosmetic
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.
- 6d ago First seen · 105 lines · 0 tokens per session scan A d5e6afb96291
comparator is an agent published in the GitHub repository gokeshenzhen/gen-tb-skill (23 stars, last pushed 1mo ago), licensed Apache-2.0. It costs nothing until one of its globs matches a file; then it loads 1,091 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
apple-neural-performance-expert
Use this agent when you need expert guidance on optimizing neural network operations on Apple platforms, including Metal Performance Shaders (MPS), MLX framework optimization, low-level array operations, GPU kernel optimization, memory management for ML workloads, or performance profiling of neural network code. This…
antenna-engineer
Reasons from gain–directivity–efficiency, Chu–Harrington bandwidth limits, and array factor through HFSS/CST/FEKO synthesis, IEEE 149-2021 NF/FF/CATR metrology, CTIA TRP/TIS/ECC OTA, and Friis link budgets while treating ground-plane truncation, active impedance in arrays, range ripple, and S₁₁≠pattern conflation as…
fabricator
3D printing and additive manufacturing specialist covering FDM, SLA, and SLS processes from model preparation through troubleshooting.
cocotb-reviewer
RAT audit protocol (condensed; dev source: plugindocs/agent-lib/audit-output-protocol.md — plugin-internal, do NOT Read it at runtime).
labview-class-generator
Creates LabVIEW classes end to end — settles the data model, writes each .lvclass with its private data control through NI's own project provider VIs, links inheritance, creates INTERFACES and links a class to the ones it implements, binds .ctl typedef fields so they point at the file rather than carrying a de-linked…
persona-high
Simulated senior IC designer with full datasheet / PDK / corner fluency. Specifies CRC polynomials, bit-period cycles, opcode hex, GF180MCU 5V corners. Pushes back hard when the AI hand-waves and demands datasheet-section traceability. Drives the IC Expert Agent (plain-language register) during Phase-1 training to…