Three-Way Review

Dispatches three independent subagents to review one or more files from distinct starting perspectives, then synthesizes their findings. Each subagent approaches the content as a reviewer with a different personality: one skeptical and critical, one generous and appreciative, one balanced with no prior lean. The goal is to surface insights that a single perspective would miss.

This skill requires subagents. If the environment does not support subagent dispatch, tell the user and stop. This skill requires textual content. Do not use this skill to review binary files.

---

Step 0: Confirm environment

Check that subagents can be dispatched. If not, explain the limitation and stop.

---

Step 1: Understand the files

Load and read each file. For each file:

• Note its type (prose, markdown, code, config, etc.)
• Note its apparent purpose
• Estimate its size

Then assess relationships between files (see "Multi-file handling" below).

Size check — do this before proceeding:

• If total content across all files exceeds 200KB, warn the user explicitly and ask whether they want to proceed. Explain that reviews may be incomplete or sampled.
• If content is under 200KB but there are many files, use a temporary staging area (e.g. a project-approved temp location, or "project memory" in environments that support memory) to avoid passing oversized context to subagents.

---

Step 2: Multi-file handling

If more than one file is provided:

1. Infer relationships first. Look for: shared imports or links, matching terminology, complementary structure (e.g., a spec and its implementation), overlapping concepts, or naming conventions that group them. Check surrounding files in the same directory if accessible; those may clarify context.

2. If the relationship is clear, proceed. Note the relationship briefly so subagents can factor it in.

3. If the relationship is unclear, ask the user before dispatching subagents:

   "Before I start, I want to confirm these files are meant to be reviewed together. Can you describe how they relate?"

4. By default, each subagent reviews all files as a set. Individual per-file reviews are only done if the user requests it.

---

Step 3: Understand the review scope

Before dispatching, determine:

• Are there guidelines? If the user mentioned specific goals, criteria, an audience, a style guide, or a particular angle (e.g., "focus on clarity", "this is for a junior audience"), pass those to all three subagents.
• Is there a requested output format? Default is a chat summary. If the user asked for file output, set write_to_file: true (see Output section).
• Any structural preferences? Default is agent-determined structure. If the user specified a format, pass it through.

---

Step 4: Dispatch subagents

Dispatch all three subagents in parallel. Pass each one:

• The full content of all files (or the staging path if using temp files)
• The relationship context (if multiple files)
• Any user-provided guidelines or goals
• The output mode (summary or file)
• Their specific reviewer persona (described below — do not use these labels in output)

Read references/subagent-personas.md for the detailed persona instructions to pass to each subagent. Read references/large-file-strategy.md for instructions to pass to subagents when any individual file is large enough to risk context limits.

If either reference file is not accessible, stop and tell the user: "The skill package appears to be incomplete — one or more required reference files are missing. Please reinstall the skill."

---

Step 5: Collect and present results

Once all three subagents return:

Default (chat summary)

Present four sections in the response:

1. Review 1 — The skeptical reviewer's findings, summarized
2. Review 2 — The generous reviewer's findings, summarized
3. Review 3 — The balanced reviewer's findings, summarized
4. Overall — A synthesis that reconciles disagreements and leans toward actionable recommendations (see Output section)

Keep each review summary concise: the key points, not a full transcript. The overall section should be the most substantive.

File output mode

Write a more comprehensive document to the requested path (or to a user/project-approved staging area if no path was given). Include:

• Full review from each subagent (not summarized, agent-determined structure unless user specified one)
• A synthesis section at the end

Then post a brief summary in chat and provide the file path.

---

Output: Overall synthesis

The synthesis is not a fourth review. It is a meta-view that:

• Notes where the three reviews agree (amplify those points, they are robust)
• Reconciles disagreements (explain the tension; recommend a direction)
• Gives concrete, prioritized recommendations
• Reflects any user-provided goals or guidelines in its framing

The synthesis should be the most actionable part of the output.