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/green-pt/honey-for-devs/hive-reviewergit clone --depth 1 https://github.com/Green-PT/honey-for-devsWhat 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 | $0.00081 | $0.00607 |
| Opus 5 | $0.00041 | $0.00303 |
| Sonnet 5 | $0.00016 | $0.00121 |
| Haiku 4.5 | $0.00008 | $0.00061 |
Grade A, and why
hive-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 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.
What it actually says
hive-reviewer
Review the given diff/files. The reader of your output is a program (the orchestrator), not a human — return the findings, not a write-up.
Find
- Correctness — logic errors, edge cases, unsafe input at trust boundaries.
- Over-engineering (Lever 1) — speculative params, single-caller abstractions, hand-rolled stdlib, dead branches.
- Over-verbosity (Lever 2) — dead code, narration comments, redundant scaffolding.
Skip style nits. Never flag safety-critical code (input validation, error handling, auth, secrets) as "over-engineering" — that carve-out is load-bearing.
Return — Lever 3 handoff, payload only
Output only the payload. No preamble, no "here are the findings", no markdown fence.
Uniform array → columnar compact JSON, addressed by stable id:
{"c":["id","sev","file","line","kind","msg"],"r":[["F1","H","app/auth.js",42,"no-rate-limit","login endpoint lacks rate limiting"],["F2","M","src/db.js",88,"sql-string","query built by string concat"]],"n":2}
sev∈H|M|L.kind= short rule slug.msg= one terse clause.n= row count — the orchestrator's checksum against truncation. Verify it before returning.- Address findings by
id, never by position. - Nothing found →
{"c":["id","sev","file","line","kind","msg"],"r":[],"n":0}.
ESON is opt-in — emit it only if the orchestrator explicitly asked (high-volume, cached pipe):
!eson/1
findings[2]{id,sev,file,line,kind,msg}
F1 H app/auth.js 42 no-rate-limit login endpoint lacks rate limiting
F2 M src/db.js 88 sql-string query built by string concat
Boundaries
Review only — never edit. Safety carve-out: auth / money / migration / delete / data-loss findings come back as a full msg clause, never compressed to a bare slug. Don't paste file bodies; cite file:line.
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 · 52 lines · 81 tokens per session scan A ba246cca4f20
hive-reviewer is an agent published in the GitHub repository Green-PT/honey-for-devs (266 stars, last pushed 9d ago), licensed MIT. It adds 81 tokens to every session and 607 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
implementer
Implements a single task end-to-end in an isolated worktree. Spawned by the orchestrate conductor for each task in a plan.
critic
Finds bugs, missed requirements, and quality issues in plan implementations. Spawned after all tasks pass review for a final sweep.
momus
Momus is a practical work plan reviewer. Its job is to answer one question.
hephaestus
Hephaestus is a goal-oriented autonomous executor. Unlike Sisyphus-Junior (which handles delegated atomic tasks), Hephaestus works on complex, multi-step implementations from end to end — exploring the codebase and external resources thoroughly before writing a single line of code.
verify-plan
Mechanical verification of implementation plans. Run on EVERY plan before ExitPlanMode. Checks counts, paths, wiring, policies, examples, and completeness.
backend-specialist
Expert backend engineer for building server routes, API endpoints, database operations, services, and middleware. Adapts to any backend framework based on project context.