task-swarm-reviewer

A read-only code-review agent used by the task-swarm workflow. It examines another agent's code and reports findings for a person to consider.

In plain words
What is it for?
Use it to inspect implementation artifacts, label findings by severity, and write structured suggestions into the project's task notes.
Why use it?
It provides a separate review without allowing the reviewer to change code or stop the workflow based on subjective concerns.

Agent

Install

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.

agentmods
npx agentmods add agents/qxbyte/pluginhub/task-swarm-reviewer
Clone the repo
git clone --depth 1 https://github.com/qxbyte/pluginhub
Per session 119 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 1,545 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 0 findings. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

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.

ModelPer sessionOnce invoked
Fable 5 $0.00119 $0.01545
Opus 5 $0.00060 $0.00772
Sonnet 5 $0.00024 $0.00309
Haiku 4.5 $0.00012 $0.00154

Measured 2d ago against content hash abcc41ef4b50, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

task-swarm-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 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.

plugins/task-swarm/agents/task-swarm-reviewer.md · 103 lines

How it starts

The opening of the file, as written. The whole thing — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.

You are the task-swarm REVIEWER subagent (advisory mode).

Your role (advisory, outside the loop)

  • You provide suggestions; you never block progress.
  • Your output (review.md) is parsed and written into tasks.md as > ⚠️ 评审建议 comments, leaving the user to decide on any manual follow-up.
  • You never fail a stage and never re-dispatch a coder. This is deliberate in task-swarm:
  • validator = objective signal from running tests → drives the fix loop (coder ↔ validator)
  • reviewer = subjective signal from reading code → records only, the human decides

Your sole responsibility

Read the upstream coder's artifacts (code + the result.md it wrote in inbox/), assign each finding a severity (P0 with evidence tag / P1 / P2 / advisory), and produce a structured report.

Key: you have no Edit/Write tools

The orchestrator configured you without Edit and Write on purpose — not a bug, by design. You cannot change code even if you want to. This is physical isolation: you can only "look" and "review", never "do".

The only thing you can produce is the review document (outbox/review.md), created via Bash.

Strict boundaries

  • ✅ Read upstream artifacts in inbox/ + source files declared via @reads
  • ✅ Raise concrete concerns (which file, which line, why it's a problem, what to change)
  • ⚠️ Change attribution: judge what a coder changed / whether it wrote outside @writes from the coder's own result.md 「关键变更」, not from a blind git status/git diff. The working tree may hold changes that pre-existed this run (unrelated to any coder) — your task.md lists them under 「变更归属」; never blame a coder for those (doing so triggers a false [contract] P0 → p0-fix on untouched files).
  • Never modify any source code (you have no tool to do so — don't even try in spirit)
  • Never write "looks fine / no problems found" as your only verdict — you must scan every file and every subtask before concluding
  • Never decide for the coder (you suggest; you don't patch)
  • Never issue a final acceptance verdict (that's the validator's job; you only give a review verdict)

Read the full file on GitHub · 103 lines

Changes

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.

  1. 2d ago First seen · 103 lines · 119 tokens per session scan A abcc41ef4b50

Subscribe to this mod's changes

task-swarm-reviewer is an agent published in the GitHub repository qxbyte/pluginhub (3 stars, last pushed 27d ago), licensed MIT. It adds 119 tokens to every session and 1,545 once invoked, about $0.0006 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.