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 commands/punt-labs/prfaq/meeting-hivegit clone --depth 1 https://github.com/punt-labs/prfaqWhat 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 | $0.00018 | $0.03873 |
| Opus 5 | $0.00009 | $0.01937 |
| Sonnet 5 | $0.00004 | $0.00775 |
| Haiku 4.5 | $0.00002 | $0.00387 |
Grade A, and why
meeting-hive 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 yesterday.
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 — 143 lines — stays where its author put it; the contents beside it link to each section on GitHub.
PR/FAQ Hive Meeting
Run an autonomous review meeting where four personas — a principal engineer, a target customer, a skeptical executive, and a builder-visionary — debate the weak spots in your PR/FAQ document and reach consensus without you moderating each decision. You review the final consensus, not each individual debate.
This is the "team meeting without the boss" variant of /prfaq:meeting.
Decision Philosophy
Amazon's Leadership Principles seek real debate, not social cohesion. Arguments must win or lose — the hive does not compromise or blend positions. When Wei says "this won't scale" and Dana says "ship it anyway," one of them wins. The loser disagrees and commits.
One-way vs two-way doors determine how the hive weighs caution vs. action:
- One-way door (irreversible): Architecture choices, public API contracts, data model commitments, third-party lock-in. On these, Wei and Alex's caution carries extra weight. The bar for overriding a feasibility or strategic concern is higher.
- Two-way door (reversible): Feature scope, positioning language, risk rating levels, FAQ framing. On these, Dana's "ship and iterate" and Priya's "does this work for me?" carry extra weight. Bias for action.
When classifying a decision, look at Wei's IRREVERSIBLE DECISIONS section — if he identifies one, it's a one-way door.
Prerequisites
This command requires Claude Code Agent Teams, which is off unless the project enables it. /prfaq:permissions sets it. If CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS is not set in the environment, tell the user to run /prfaq:permissions, or to add this to the project's .claude/settings.json by hand:
{
"env": { "CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS": "1" }
}
Then restart Claude Code. Do not proceed without it — use /prfaq:meeting for manual moderation instead.
Steps
- Find the document. If
$ARGUMENTSspecifies a path, use it. Otherwise, search forprfaq.texin the project root using Glob. If no document exists, tell the user to run/prfaqfirst.
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.
- yesterday First seen · 143 lines · 18 tokens per session scan A 6a12e20fc38c
meeting-hive is a command published in the GitHub repository punt-labs/prfaq (25 stars, last pushed yesterday), licensed MIT. It adds 18 tokens to every session and 3,873 once invoked, about $0.0001 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 commands, from other repositories
oracle-dev
Generate property-based test harness using Lean 4 model as oracle.
refine-dev
Verify code refines Z specification via abstraction function and commutativity checks.
refine
Verify code refines Z specification via abstraction function and commutativity checks.
contracts-dev
Generate runtime contracts (preconditions, postconditions, invariants) from a Z specification.
code2model-dev
Create a Z specification for stateful entities in a system.
partition-dev
Derive test cases from Z specification using TTF testing tactics.