Guide · Agent skill

How to use the Agent Transcript Publisher agent skill

Updated 2026-11-02 · written with AI from the project's own documentation and checked against our catalogue data

The Agent Transcript Publisher is an agent skill that lets an AI coding agent turn a local session log into a condensed Markdown transcript, review it, and, only when you authorize it, insert it into a GitHub pull request or issue. It is aimed at maintainers and developers who want a reviewed record of how a PR was produced. After this guide you will know how to install the skill, how to phrase requests so it triggers, and what the safety steps require before anything is published.

Agent Transcript Publishersteipete/agent-scripts · MIT Download the skill

Quick start

1Download and unzip the skill. Download the skill ZIP from this page and unzip it into ~/.claude/skills/ or your project's .claude/skills/ folder.
2Confirm the prerequisites. Make sure you can run the bundled agent-transcript script and that local Codex, Claude, Pi or OpenClaw logs exist on your machine.
3Ask explicitly for a transcript. Request a transcript or preview by name, stating the PR or issue if relevant. Ordinary PR work will not trigger it.
4Find and render the session. Run the find command to locate candidate logs, then render the chosen session file to Markdown in /tmp.
5Trim and inspect the Markdown. Remove system prompts, reasoning, raw tool output, local paths, credentials and auth URLs, then read the whole result.
6Authorize and insert. Show the trimmed text for review. Insert it into the named PR or issue only after you explicitly authorize that publication.

What an agent skill is and what this one does

An agent skill is a folder of instructions that an AI agent loads when a request matches the skill's purpose. This skill is described as handling requested GitHub PR and issue agent transcripts: it redacts, trims, previews and inserts them safely.

The skill is designed to act only when you explicitly ask for a transcript or preview. It reads local Codex, Claude, Pi and OpenClaw session logs, renders them to Markdown with a bundled helper script, and then has the agent manually trim the result to the task-relevant parts. Trimming removes system and developer prompts, reasoning, raw tool output, environment data, local paths, credentials, browser or cookie details, and auth URLs.

Triggers only on an explicit request for a transcript or preview.
Discovery and rendering use no network, according to the skill.
Shows the trimmed Markdown for review before any insertion.
Inserts into a PR or issue only when your authorization names that publication.

Install the skill from the ZIP download

Download the ZIP file for the skill from this page. Then install it in one of two ways. For a personal setup, unzip it into ~/.claude/skills/. For a single project, unzip it into that project's .claude/skills/ folder.

If you use the Claude apps instead, open Settings, go to Capabilities, and upload the skill under Skills.

Personal: unzip into ~/.claude/skills/
Project: unzip into the project's .claude/skills/ folder
Claude apps: Settings → Capabilities → Skills → upload

Check what the skill needs before you run it

The skill runs a bundled helper script at skills/agent-transcript/scripts/agent-transcript, so you need a Python or shell environment that can run it. The source for that script was not included in the files reviewed, so inspect it yourself before relying on it.

You also need local session logs from Codex, Claude, Pi or OpenClaw. Publishing additionally requires GitHub access with permission to edit the target PR or issue. The skill does not use the network for discovery or rendering.

Python or shell able to run skills/agent-transcript/scripts/agent-transcript
Local Codex, Claude, Pi or OpenClaw session logs
GitHub access with edit permission on the target PR or issue, only if publishing

Ask for a transcript with a natural request

The skill expects an explicit request. Say plainly that you want a transcript or a preview of the session, and name the PR or issue if you want it inserted. A request for a preview does not by itself permit publishing.

Ordinary PR work does not trigger the skill. The skill is written to stay out of routine PR creation and editing unless a transcript is explicitly requested.

"Make a transcript of this session and show me the trimmed version before anything is posted."
"Preview a transcript of the Codex session that produced this PR."
"Add the reviewed transcript to PR #123 in a collapsed section." (publishing authorized for that PR only)

How the helper finds and renders a session

The skill's workflow uses two helper commands. The find command searches local logs for candidates matching your PR title or branch. The render command converts one session file to Markdown. Treat find results as candidates, not proof that a session covers the work.

By default find scans the newest 400 matching logs. The --max-files flag widens discovery. After rendering, the agent must trim the Markdown before showing, previewing or inserting it.

find scans the newest 400 matching logs by default; use --max-files N to widen the search
Treat matches as candidates, not proof of scope
Trim the rendered file before any display or insertion

skills/agent-transcript/scripts/agent-transcript find \
  --query "$PR_TITLE $BRANCH_OR_PR_URL" --cwd "$PWD" --since-days 14
skills/agent-transcript/scripts/agent-transcript render \
  --session "$SESSION_JSONL" --out /tmp/agent-transcript.md

Trim, review and insert only what you approved

Trimming is the key safety step. The agent keeps only task-relevant user prompts, visible decisions, terse tool summaries and proof outcomes, using the PR or issue goal, branch and changed files to drop unrelated turns. The skill states that helper redaction is not sufficient disclosure review, so you must inspect the result yourself. If sensitive content cannot be resolved, the transcript should be omitted.

The helper's preview, html and append-body modes render the session again from the original log, so they can bring back turns you removed. Do not treat their untrimmed output as approved. Insertion uses only the inspected, scoped text, and existing transcript markers are replaced rather than duplicated. The collapsed details section is kept.

Keep only the trimmed Markdown you have read
Do not use preview, html or append-body output as the final artifact
If no safe match exists, the PR work continues without a transcript and the omission is explained

What people use it for

Attaching a condensed, reviewed record of how a maintainer's PR was produced to the PR itself.
Sharing the relevant part of a Codex or Claude session with an issue thread without exposing unrelated work.
Previewing a trimmed transcript so a team can review it before anything is published.
Replacing an earlier transcript block in a PR body instead of adding a duplicate.

Frequently asked questions

Is the Agent Transcript Publisher skill free to use?

The skill is published under the MIT licence, as stated on this page. Check the licence file in the repository for the full terms before reuse.

What do I need to run the agent-transcript skill?

You need a Python or shell environment that can run the bundled agent-transcript script, local Codex, Claude, Pi or OpenClaw session logs, and GitHub access with edit permission if you want to publish. The skill does not use the network for discovery or rendering.

Where do I install the agent skill?

Unzip the downloaded ZIP into ~/.claude/skills/ for personal use or into a project's .claude/skills/ folder. In the Claude apps you can instead upload it under Settings, Capabilities, Skills.

Will the skill post transcripts to my PRs automatically?

No. The skill is written to insert text only after you clearly authorize publishing that transcript to a specific PR or issue. A preview request alone does not authorize publication.

Is the helper's redaction enough to make a transcript public-safe?

No. The skill says helper redaction is not sufficient disclosure review. You need to trim and read the Markdown yourself, especially for public repositories where logs may contain secrets.

What should I use instead if I want automatic transcript posting?

This skill is built around human review and deliberately avoids automatic posting. If you need automation without review, this skill is the wrong fit, and you would need a different tool; the source does not name alternatives.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

This skill suits maintainers and developers who want a reviewed, condensed transcript of a session attached to a PR or issue, and teams that require human review before publication. It is a poor fit for routine PR work, for public repositories where you cannot verify that secrets were removed, or for anyone who will not read the rendered Markdown before it goes live.

Download the skill