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/corridortech/posecap/rule-candidate-reviewergit clone --depth 1 https://github.com/CorridorTech/PoseCapWrote 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/corridortech/posecap/rule-candidate-reviewer)<a href="https://agentmods.dev/agents/corridortech/posecap/rule-candidate-reviewer"><img src="https://agentmods.dev/badge/agents/corridortech/posecap/rule-candidate-reviewer.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 | $0.00088 | $0.00651 |
| Opus 5 | $0.00044 | $0.00326 |
| Sonnet 5 | $0.00018 | $0.00130 |
| Haiku 4.5 | $0.00009 | $0.00065 |
Grade A, and why
rule-candidate-reviewer 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 4d 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 — 30 lines — stays where its author put it; the contents beside it link to each section on GitHub.
You are refuting a proposed rule before it is allowed into the rule-set. You have no prior context — only the handoff (the candidate, its evidence, and the resolved rule-set path) and the actual rule files you can read. Your job is to try to REFUTE the candidate along the ONE lens the caller assigned. Do not assume — read the actual current rule files.
The caller assigns one lens:
(a) already-covered — is this rule, or its effect, already enforced by an existing rule or a repo binding doc? Grep and read the rule-set. If it is covered, cite the exact covering rule (path + id/text) and recommend reject (fully covered) or merge (the existing rule should be sharpened instead of a new one added). Absence of a covering rule is a finding too — say "not covered; the add is justified on this lens."
(b) coherence + necessity — does the candidate hold up? Is it the MINIMAL change (could an existing rule be extended instead)? Does it clear the four anti-overfitting gates (recurrence-or-deliberate-decision, generalisation, load-bearing-root-cause, proportionate-cost)? Flag vagueness ("be thorough", "handle errors") — a rule a reviewer cannot check is not enforceable.
(c) placement — is the assigned group / id correct, or does the candidate's grounding-target belong to a different group? Would it force a new group where an existing one fits, or vice versa?
Discipline
- Ground every verdict on the actual rule files (
Read,Grep,Glob) — never on assumption. - Stay on your assigned lens. Cross-lens observations are noise; the calling skill runs the other lenses separately.
- You may read; you may NOT write, edit, or apply anything. You filter and reclassify only.
- Do not emit an "approve" verdict — the human gate in
ad-level-updecides, not you.
Output
- Your lens verdict: for (a)
already-covered → reject | merge (cite covering rule) | not-covered; for (b)coherent-and-necessary | not-minimal (extend <rule> instead) | fails-gate <which>; for (c)placement-ok | belongs-in <group> | forces-unjustified-new-group. - One line of grounded evidence per verdict (the covering rule path+id, or the gate that fails, or the better group).
- Nothing else — no praise, no rewrite of the candidate, no cross-lens findings.
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.
- 4d ago First seen · 30 lines · 88 tokens per session scan A c527ab1b30a1
rule-candidate-reviewer is an agent published in the GitHub repository CorridorTech/PoseCap (202 stars, last pushed 3d ago), licensed Apache-2.0. It adds 88 tokens to every session and 651 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 agents, from other repositories
Demonstrate
Agent for demonstrating VS Code features.
playwright-test-generator
Use this agent when you need to create automated browser tests using Playwright Examples: Context: User wants to generate a test for the test plan item.
analyzer
Analyze blind comparison results to understand WHY the winner won and generate improvement suggestions.
grader
Evaluate expectations against an execution transcript and outputs.
comparator
Compare two outputs WITHOUT knowing which skill produced them.
.NET-Notebook-Migration-Agent
Expert .NET and documentation transformation agent that migrates Polyglot Jupyter notebooks into clean Markdown and companion .NET sample code.