Borrowing it
Nothing to install: this file belongs to sean-galloway/RTLDesignSherpa. 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/sean-galloway/RTLDesignSherpa/main/.claude/skills/review-rounds/SKILL.mdgit clone --depth 1 https://github.com/sean-galloway/RTLDesignSherpaWrote 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/sean-galloway/rtldesignsherpa/review-rounds)<a href="https://agentmods.dev/skills/sean-galloway/rtldesignsherpa/review-rounds"><img src="https://agentmods.dev/badge/skills/sean-galloway/rtldesignsherpa/review-rounds/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/sean-galloway/rtldesignsherpa/review-rounds"><img src="https://agentmods.dev/badge/skills/sean-galloway/rtldesignsherpa/review-rounds.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 | $0.00079 | $0.00610 |
| Opus 5 | $0.00039 | $0.00305 |
| Sonnet 5 | $0.00016 | $0.00122 |
| Haiku 4.5 | $0.00008 | $0.00061 |
Grade A, and why
review-rounds 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 11d 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.
What it actually says
review-rounds
READ FIRST: vault/handbook/INDEX.md (the handbook is the repo's memory; this skill is the signpost). Canonical: vault/handbook/authoring/kimi-review-rounds.md - the seven rules, both round modes, and the endpoint config.
Four that bite hardest when triaging findings:
- The second-model adjudication pass is ADVISORY. A REFUTED verdict never
drops a finding on its own -- 4 of the ~7 it has issued were wrong, against
a reviewer FP rate of 2 in 72. Measure the extractor first
(
verify_findings.py --dry-runprints the located-quote share); a blind verdict is not evidence in either direction. - A finding that reads like a doc nit can be a real RTL defect (and vice versa). Read the whole finding, not the headline; triage doc-fix vs RTL-fix per finding.
- Integration status is MEASURED against the tree, never inferred from commit history. A "reconcile docs with the RTL" commit is not evidence a round was applied - one landed six hours before a round that then found 70 confirmed defects.
- Verify a fix with a CLEAN REBUILD and mutation-check the test: a stale sim_build passes against the old RTL, and a test whose stimulus cannot expose the bug passes against the broken RTL. Revert, confirm RED, restore.
Voice pass: [[humanization-voice]]. Off-workstation runs: [[cloud-sandbox]].
Scripts: bin/build_review_bundle.py (rebuild ALL units, always) then bin/review/run_batch.py {qc|humanize} (serial, never overwrites a round).
Direct-mode runbook (off the litellm proxy) is in the handbook note: model is ALWAYS kimi-k3; the Moonshot key loads inline from an out-of-repo secrets store and NEVER enters the repo (not even its path); bundle + results live outside the working tree. The bundle misses index/readme/overview (only what book*_index links) -- add an _meta unit for a send-ALL-md pass.
The handbook root is vault/handbook/INDEX.md - design/, dv/, fpga/, authoring/ areas, atomic notes, wikilinked. When you learn a durable lesson in this domain, ADD IT TO THE HANDBOOK NOTE, not to this skill.
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.
- 11d ago First seen · 43 lines · 79 tokens per session scan A 2c1ac564096b
review-rounds is a skill published in the GitHub repository sean-galloway/RTLDesignSherpa (23 stars, last pushed today), licensed MIT. It adds 79 tokens to every session and 610 once invoked, about $0.0004 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 2026-08-30.
Other skills, from other repositories
arch-review
Read-only architecture review of RTL vs uArch spec with area/timing/power tradeoffs. Use for post-RTL architecture sign-off or suspected spec mismatch.
cdc-tool-profiles
Internal reference: cdc tool profiles (agent-loaded; do not invoke).
code-review-policy
Internal reference: code review policy (agent-loaded; do not invoke).
refactor-classification-policy
Internal reference: refactor classification policy (agent-loaded; do not invoke).
p2-arch-design-policy
Internal reference: p2 arch design policy (agent-loaded; do not invoke).
codex-cross-review
Cross-review with Codex CLI as independent 2nd reviewer; finding exchange until consensus or escalation. At phase boundaries or on 'codex review'.