What is watch-session?

A workflow for developing or improving a coding-agent skill by watching another Claude session and then testing revisions in a separate trial session. It uses herdr to read a session and send prompts when requested.

It watches a session, starts from an existing skill or creates a new one, edits the skill, and proves revisions through trials.

It captures how a person or agent actually completes a task, then checks whether a new skill can reproduce that result without guidance.

watch-session — Skill for Claude Code by bendrucker/claude

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.

Any agent

npx skills add bendrucker/claude --skill watch-session

Clone the repo

git clone --depth 1 https://github.com/bendrucker/claude

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

README.md

[![agentmods](https://agentmods.dev/badge/skills/bendrucker/claude/watch-session/github.svg)](https://agentmods.dev/skills/bendrucker/claude/watch-session)

Your own site

<a href="https://agentmods.dev/skills/bendrucker/claude/watch-session"><img src="https://agentmods.dev/badge/skills/bendrucker/claude/watch-session/github.svg" alt="Measured on agentmods" height="20"></a>

Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.

Your own site · 80×15

<a href="https://agentmods.dev/skills/bendrucker/claude/watch-session"><img src="https://agentmods.dev/badge/skills/bendrucker/claude/watch-session.svg" alt="Reviewed on agentmods" width="80" height="20"></a>

Per session 60 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.

When invoked 1,775 The whole file, excluding the scripts and references it only reads on demand.

Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.

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.

Model	Per session	Once invoked
Fable 5.1	$0.00060	$0.01775
Opus 5.5	$0.00024	$0.00710
Sonnet 5.5	$0.00012	$0.00355
Haiku 4.5	$0.00006	$0.00178

Measured 10d ago against content hash 38cf0f9e40b9, method: parsed. Prices are Anthropic first-party input rates as of 2026-10-07, from the pricing page.

Security

Grade A, and why

watch-session 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 10d ago.

The scan reads SKILL.md. This mod also ships 7 executable files (scripts/herdr.test.ts, scripts/herdr.ts, scripts/transcript.test.ts, …), listed below but not scanned — reading those needs a real analyzer, not pattern matching.

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.

user/skills/watch-session/SKILL.md · 98 lines

How it starts

The opening of the file, as written. The whole thing — 98 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Watch Session

Develop a skill by watching the user work in another Claude session, then proving each revision in a trial. The goal is a skill that lets a fresh session reach the user's result without the user steering it. Watching tells you what to try. A trial settles whether it works.

The user's session belongs to the user. You read its transcript, and type into it only when the user asks, such as to start the run they want watched: herdr agent prompt <pane> "<prompt>" once herdr agent wait <pane> returns. Omit --until, because a pane in an unseen tab rests at done and the default matches it. Trials run in a pane you open and own.

Arguments

<pane-id | session-uuid>: the user's session. A herdr pane id (wMQ:p1) resolves to the Claude session in that pane and follows it across /clear. A session UUID watches that transcript directly.
--skill <dir>: start from an existing skill and edit it in place. Without it or --learn, take the directory from the first skillDirs in a turn digest and confirm it with the user before the first edit.
--learn <name>: start from nothing and draft a skill named <name> in tmp/watch-session/<uuid>/.claude/skills/<name>/.
--prompt <text>: the prompt a trial sends. Default: the user's first typed prompt of the task.
--every <duration>: batch interval while observing a hand-done task. Default 10m with --learn, per turn otherwise.

Load plugin:skill and prompting:prompting before the first edit.

State

Keep everything for one watch in tmp/watch-session/ under your working directory, keyed by the user's session UUID from the watching event:

<uuid>.json: the script's read offset, written through --state-dir.

<uuid>.md: your log. Append one entry per revision, trial, or open question:

turn 88213-131040  re-ran the full benchmark after a no-op edit
  -> SKILL.md: skip the bench when the diff touches no measured path
trial 2  reached the user's end state, 1 wasted bench run

Read the full file on GitHub · 98 lines

Files

What ships with it

8 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.

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.

10d ago Changed · -2 tokens per session 38cf0f9e40b9
18d ago First seen · 98 lines · 62 tokens per session scan A 7184b5d4cbab

Subscribe to this mod's changes

watch-session is a skill published in the GitHub repository bendrucker/claude (18 stars, last pushed today), licensed MIT. It adds 60 tokens to every session and 1,775 once invoked, about $0.0002 per session on Opus 5.5. 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 2026-09-29.

Building agents Claude Code Testing claude-code-plugin

live-preview

Mid-build visual verification loop. Takes screenshots of components during construction, not just after. Catches visual regressions and invisible features before they compound. Requires Playwright or similar screenshot tool.

SethGammon/Citadel · 40 tokens

brooks-sweep

Full-sweep mode: runs a unified analysis across all quality dimensions — code decay, architecture, tech debt, and test quality — then applies fixes directly to the codebase. Safe changes are auto-applied; risky changes are confirmed before execution. Drawing on twelve classic engineering books. Triggers when: user…

hyhmrright/brooks-lint · 178 tokens

watch-session