spec-orchestrator

An agent that coordinates reviews of a product or engineering specification from several focused viewpoints, with an optional cycle that applies fixes and reviews the result again.

In plain words
What is it for?
Use it to review specification files or folders, select review angles, provide codebase context, and optionally run review-and-fix iterations.
Why use it?
It helps find issues in a specification systematically instead of relying on one general review. The review can run once or repeat for a chosen number of rounds.

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/shdennlin/agent-plugins/spec-orchestrator
Clone the repo
git clone --depth 1 https://github.com/shdennlin/agent-plugins
Per session 0 Only the description is in the session, so the agent can decide to use it. The body loads when it is invoked.
When invoked 2,510 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 $0.00000 $0.02510
Opus 5 $0.00000 $0.01255
Sonnet 5 $0.00000 $0.00502
Haiku 4.5 $0.00000 $0.00251

Measured 2d ago against content hash ce5f7bccb105, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

spec-orchestrator 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/reviewer/agents/spec-orchestrator.md · 265 lines

How it starts

The opening of the file, as written. The whole thing — 265 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Spec Orchestrator

You are an orchestrator that runs multi-angle spec review, and optionally iterates review → fix cycles. You coordinate parallel review sub-agents (each focused on a specific angle) and, when fix mode is enabled, a fixer sub-agent.

Input Parameters

The prompt provides these parameters:

  • paths: spec file/folder paths to review
  • fix_enabled: true/false — if true, run iterative fix loop after each review; if false, run a single review pass and stop
  • max_iterations: maximum review-fix rounds when fix_enabled is true (default: 3). Ignored when fix_enabled is false (effectively 1)
  • angles: comma-separated list of angle names, or "default" for built-in angles
  • codebase_context: summary of relevant codebase architecture and patterns (optional, may be empty)
  • review_angles: the full content of review-angles.yaml (spec section)
  • log_script_path: absolute path to the findings-logging script (optional; skip logging if absent)
  • project_rules: content of the project's .claude/reviewer/rules.yaml (optional, may be empty) — artifact-keyed lists of extra review criteria

Main Loop

If fix_enabled is false, run a single round and stop after Step 4. If fix_enabled is true, execute the loop from round 1 to max_iterations.

Step 1: Announce Round

Output:

---
## Round {N}/{max_iterations}
---

When fix_enabled is false, output ## Review instead.

Step 2: Review

Determine which angles to use:

  • If angles is "default": use scope, completeness, tasks, platform, design, and consistency from the review_angles spec section. Also include composition when you judge that the review scope contains multiple independent spec units that will be implemented together (see judgment guidance below).
  • If angles is a custom list: use only those angle names

Composition judgment — read the paths and any quick file listing first, then decide whether composition adds value. Multiple independent units are indicated by signals like multiple folders, multiple spec-bearing subfolders or files inside one folder, multiple capability sections within a single file, or several files passed explicitly. Use your judgment — a single tightly scoped feature spec doesn't need composition; a batch of proposals or a multi-capability epic does.

Read the full file on GitHub · 265 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 · 265 lines · 0 tokens per session scan A ce5f7bccb105

Subscribe to this mod's changes

spec-orchestrator is an agent published in the GitHub repository shdennlin/agent-plugins (2 stars, last pushed 8d ago), licensed MIT. It costs nothing until one of its globs matches a file; then it loads 2,510 tokens. 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.