adversarial-authoring

A review process in which one configured agent writes a draft and another challenges it before the final version is produced. It is intended for instructions that call for adversarial authoring, cross-model review, or a model council.

In plain words
What is it for?
Use it when an artifact needs a draft, an independent challenge, and a recorded review process before completion.
Why use it?
It reduces the risk that one perspective misses errors or requirements. The primary agent then weighs the review against the original task and project rules.

Skill for Claude CodeCodex

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 skills/intent-driven-dev/intent-driven-template/adversarial-authoring
Any agent
npx skills add intent-driven-dev/intent-driven-template --skill adversarial-authoring
Clone the repo
git clone --depth 1 https://github.com/intent-driven-dev/intent-driven-template

Made for: Claude Code, Codex.

Per session 65 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 839 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 $0.00065 $0.00839
Opus 5 $0.00032 $0.00419
Sonnet 5 $0.00013 $0.00168
Haiku 4.5 $0.00006 $0.00084

Measured 2d ago against content hash c62e94322478, method: parsed. Prices are Anthropic first-party input rates as of 2026-08-30, from the pricing page.

Security

Grade A, and why

adversarial-authoring 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.

.opencode/skills/adversarial-authoring/SKILL.md · 88 lines

How it starts

The opening of the file, as written. The whole thing — 88 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Adversarial Authoring

Use this skill when an artifact rule, user request, or instruction says to use adversarial authoring, a model council, cross-model review, or a mixture of models.

The goal is to reduce single-perspective bias by having one configured subagent author the work and another configured subagent challenge it before the primary agent writes the final artifact.

Required Workflow

  1. Identify the artifact or task being authored, including its target output path.
  2. Gather the full task context before calling subagents:
    • User request
    • Relevant OpenSpec instruction JSON
    • Artifact template
    • Project context
    • Artifact-specific rules
    • Dependency artifacts already completed
  3. Start the authoring round with the adversarial-author subagent.
  4. Send the author draft and the original task context to the adversarial-reviewer subagent.
  5. Reconcile the review yourself as the primary agent:
    • Accept corrections that improve correctness, clarity, requirements coverage, or risk handling.
    • Reject suggestions that conflict with the user request, artifact template, OpenSpec instructions, or project rules.
    • Ask the user only if the disagreement changes product intent or cannot be resolved from available context.
  6. Write the final artifact to the requested output path.
  7. Write council notes beside the artifact using the naming convention below.

Subagents

Use these hard-coded opencode subagents:

  • adversarial-author: authors the initial draft.
  • adversarial-reviewer: reviews and challenges the draft.

The specific model providers are intentionally defined only in the agent files, so changing model families only requires updating each agent's model frontmatter.

Do not substitute other agents unless the user explicitly requests it.

Council Notes

For every artifact authored through this skill, create or update a sibling council-notes file.

Naming convention:

  • proposal.md -> proposal.council.md
  • design.md -> design.council.md
  • tasks.md -> tasks.council.md
  • spec.md -> spec.council.md

Read the full file on GitHub · 88 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 · 88 lines · 65 tokens per session scan A c62e94322478

Subscribe to this mod's changes

adversarial-authoring is a skill published in the GitHub repository intent-driven-dev/intent-driven-template (121 stars, last pushed 9d ago), licensed MIT. It adds 65 tokens to every session and 839 once invoked, about $0.0003 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.