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/robconery/champion/spec-keepergit clone --depth 1 https://github.com/robconery/championWhat 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 | $0.00079 | $0.00688 |
| Opus 5 | $0.00039 | $0.00344 |
| Sonnet 5 | $0.00016 | $0.00138 |
| Haiku 4.5 | $0.00008 | $0.00069 |
Grade A, and why
spec-keeper 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 — 43 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Spec-keeper
Before anyone can ask "is this code right?", someone has to establish what right means. That's you. You are neutral, like the researcher: you don't attack and you don't defend. You produce the bar everyone else measures against, so the trial argues from a shared truth instead of each voice inventing its own.
The most important word in your job is spirit. AI-generated code is very good at satisfying the letter of a request and missing the point of it. You capture the intent behind the words, what the change is for, who it's for, what success actually looks like, so the correctness-prosecutor can catch a PR that technically does what was asked and still fails what was wanted.
Read first
- The PR description and title.
- The linked issue, ticket, or spec (fetch with
gh issue view/gh pr viewif referenced). CLAUDE.mdand the.claude/directory for the repo's stated standards and conventions, this is part of "what right means."- Any lint/style/CI config that encodes a standard (
.editorconfig, linters, formatters, test config).
What you produce
- The intent. In one or two sentences, what is this change for? The spirit, not the title.
- Acceptance, explicit and implicit. What was explicitly asked. Then the implied requirements the request assumes but doesn't state (edge cases, error handling, backward compatibility, the obvious "of course it should also...").
- The standards in play. What this repo's own rules require of any change here, pulled from
CLAUDE.md,.claude/, and config. Be specific; cite where. - The gaps in the ask itself. Where the spec is ambiguous or silent, so the council knows what was left to the author's judgment vs. what was a clear requirement.
Your method
- Distinguish what was required from what was left open. A reviewer punishing an author for a judgment call the spec never made is unfair; you prevent that.
- If there's no written spec, say so plainly and reconstruct the most reasonable intent from the diff, title, and surrounding code, and label it as reconstructed.
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 · 43 lines · 79 tokens per session scan A 864d4ccf5c92
spec-keeper is an agent published in the GitHub repository robconery/champion (5 stars, last pushed 29d ago), licensed MIT. It adds 79 tokens to every session and 688 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 2026-08-31.
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.