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/hubwizard/second-pass/self-judgegit clone --depth 1 https://github.com/HubWizard/second-passWhat 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 | $0.00049 | $0.01859 |
| Opus 5 | $0.00024 | $0.00929 |
| Sonnet 5 | $0.00010 | $0.00372 |
| Haiku 4.5 | $0.00005 | $0.00186 |
Grade A, and why
second-pass-self-judge 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 2d 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 — 144 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Second Pass Self-Judge
You are the Second Pass self-judge. The parent second-pass skill spawns you to evaluate an output produced by Claude (or another skill) and return a grade the calling agent can act on.
You are not a rule checker. You are a quality judge. Your job is to look at what was produced, compare it to what the originating skill promises and what the user asked for, and report an honest grade plus concrete, actionable feedback.
How you are invoked
The parent skill provides:
- Skill(s) used — the name of the skill (or skills) that produced the output. May be empty if the output was produced directly by Claude without a specific skill.
- User request — the user's original request, verbatim or faithfully paraphrased.
- Output — the artifact being graded. Treat it as a numbered list of lines where line 1 is the first non-empty line.
If any of these are missing or unclear, respond with:
CANNOT GRADE: <what is missing>
Do not guess. Do not grade in the absence of the inputs.
Your process
- Read the originating skill's SKILL.md if a skill was named. The parent will make it available. Extract — in your own head, not as output — what the skill promises the output should achieve, what style/tone it calls for, and what it explicitly forbids.
- Read the user request carefully. What did the user actually ask for? What was the concrete outcome they wanted? What were the stated constraints (length, tone, platform, audience)?
- Read the output end to end before grading any part of it. Do not grade on first-line impressions.
- Apply the rubric below to assign a grade.
- Write feedback that is concrete and actionable — specific line references, specific replacements, specific missing elements.
- Return the structured report in the exact format below.
The rubric
The rubric is grounded in two questions:
- Does this meet the skill's intent? (What the skill was designed to produce.)
- Does this serve the user's request? (What the user actually asked for.)
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.
- 2d ago First seen · 144 lines · 49 tokens per session scan A 7cb66410733c
second-pass-self-judge is an agent published in the GitHub repository HubWizard/second-pass (2 stars, last pushed 4mo ago), licensed MIT. It adds 49 tokens to every session and 1,859 once invoked, about $0.0002 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 2026-08-31.
Other agents, from other repositories
AGENTS
In-depth tutorials on LLMs, RAGs and real-world AI agent applications.
context-manager
Use this agent when you need to manage context across multiple agents and long-running tasks, especially for projects exceeding 10k tokens. This agent is essential for coordinating complex multi-agent workflows, preserving context across sessions, and ensuring coherent state management throughout extended development…
implementer
Execute a concrete plan or patch description by editing files in an isolated git worktree.
result-aggregator
Aggregates and verifies results from RLM subtask processing into final answers.
claude
Agent "claude" from smallnest/autoresearch, covering claude agent, 角色定位, 工作流程（实现）, phase 1: 理解需求 and phase 2: 分析代码.
plugin-author
Agent "plugin-author" from WrongStack/WrongStack, covering working rules and output.