review-session

name	review-session
description	Use when the user wants to review an opencode session for quality — analyzes a specified session (or the most recent non-active one) for prompting effectiveness, agent performance, and environment gaps, producing actionable recommendations

Review Session

Review a single opencode session transcript for quality and produce actionable findings.

Prerequisites

• The opencode CLI must be available on PATH (provides opencode session list and opencode export).
• The ed3d-conversation-reviewer agent and the writing-opencode-directives skill must be available (the reviewer loads that skill).
• This skill ships scripts/reduce-transcript.py (its absolute path is shown in the file list when this skill loads).

Invocation

The user may ask to:

• review "this session" or the most recent session — pick the most recent non-active session, or ask which one (see step 1)
• review a specific session — identified by session ID (ses_…) or by title

Steps

1. Determine the session to review

List the current project's sessions:

opencode session list

Session IDs match ses_[A-Za-z0-9]+.

Never export the active session — a session still being written can truncate mid-write. The active session is the most recently updated one (its title matches the running conversation).

• If the user named a session (by ID or title), use that one.
• If the user asked for "this session" or didn't specify: take the most recent non-active session, or ask the user which session they want if it's ambiguous.

2. Export and reduce the transcript

Export the session, then run scripts/reduce-transcript.py (shipped with this skill; its absolute path is shown in the file list when this skill loads) to produce a token-efficient version:

mkdir -p /tmp/session-review
opencode export <sessionID> > /tmp/session-review/session.json
python3 scripts/reduce-transcript.py /tmp/session-review/session.json /tmp/session-review/reduced.txt

Tell the user you're reducing the transcript for analysis.

3. Dispatch the reviewer

Create a temporary output path for findings:

/tmp/session-review/findings.md

Dispatch the ed3d-conversation-reviewer agent with:

task:
  subagent_type: ed3d-conversation-reviewer
  description: Review session transcript
  prompt: |
    Review the reduced opencode session transcript.

    Transcript path: /tmp/session-review/reduced.txt
    Write your findings to: /tmp/session-review/findings.md

    Read the transcript, analyze it, and write your findings following your output format.

4. Present findings

Once the reviewer completes, read /tmp/session-review/findings.md and present the findings to the user.

Present the full findings — do not summarize or truncate. The reviewer has already calibrated the length to be proportional to what's interesting.