watercooler-onboarding

watercooler-onboarding is a skill for Claude Code from mostlyharmless-ai/watercooler. It costs 79 tokens per session (12,069 once invoked), scanned A, original, Apache-2.0.

A repository-onboarding skill that examines a codebase, its documentation, continuous-integration setup, Git history, and existing Watercooler discussion threads. Watercooler is the memory system it uses to save durable, source-backed notes for future agents.

In plain words
What is it for?
Use it when first entering a repository, refreshing its saved context, previewing proposed notes, or updating CLAUDE.md or AGENTS.md after onboarding.
Why use it?
It reduces the time needed for a new agent to understand a repository and preserves useful project context for later work.

Skill for Claude Code

Written for Claude Code: allowed-tools in frontmatter. Also seen: mentions CLAUDE.md; mentions subagents; mentions Claude Code.

Part of the watercooler plugin — 7 skills, 1 MCP server shipped together

Good fit Use it when first entering a repository, refreshing its saved context, previewing proposed notes, or updating CLAUDE.md or AGENTS.md after onboarding.

Compare 6 skills from other repositories ↓
Install with agentmods
npx agentmods add skills/mostlyharmless-ai/watercooler/watercooler-onboarding
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 mostlyharmless-ai/watercooler --skill watercooler-onboarding
Clone the repo
git clone --depth 1 https://github.com/mostlyharmless-ai/watercooler

Made for: Claude Code.

Or install watercooler, the plugin that ships this one along with the rest of its 7 skills, 1 MCP server.

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.

agentmods badge for watercooler-onboarding

README.md
[![agentmods](https://agentmods.dev/badge/skills/mostlyharmless-ai/watercooler/watercooler-onboarding/github.svg)](https://agentmods.dev/skills/mostlyharmless-ai/watercooler/watercooler-onboarding)
Your own site
<a href="https://agentmods.dev/skills/mostlyharmless-ai/watercooler/watercooler-onboarding"><img src="https://agentmods.dev/badge/skills/mostlyharmless-ai/watercooler/watercooler-onboarding/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.

agentmods 80×15 button for watercooler-onboarding

Your own site · 80×15
<a href="https://agentmods.dev/skills/mostlyharmless-ai/watercooler/watercooler-onboarding"><img src="https://agentmods.dev/badge/skills/mostlyharmless-ai/watercooler/watercooler-onboarding.svg" alt="Reviewed on agentmods" width="80" height="20"></a>
Per session 79 Skills are progressive disclosure: only the name and description are preloaded; the body loads when the skill is used.
When invoked 12,069 The whole file, excluding the scripts and references it only reads on demand.
Security scan A 1 finding. 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.

ModelPer sessionOnce invoked
Fable 5.1 $0.00079 $0.12069
Opus 5 $0.00039 $0.06034
Sonnet 5 $0.00016 $0.02414
Haiku 4.5 $0.00008 $0.01207

Measured 10d ago against content hash 50c84ec5ae05, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-10, from the pricing page.

Security

Grade A, and why

watercooler-onboarding scanned grade A with 1 finding 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.

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.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

`fetch-papers` curl safety, extract to markdown via the `pdf-to-md` per-PDF Task subagent
plugins/claude/watercooler/skills/watercooler-onboarding/SKILL.md · 868 lines

How it starts

The opening of the file, as written. The whole thing — 868 lines — stays where its author put it; the contents beside it link to each section on GitHub.

Watercooler Repository Bootstrap

Create durable Watercooler seed context for a repository.

Arguments: $ARGUMENTS

Default behavior is Bootstrap: inspect the repository and write a bounded set of Watercooler entries. Read-only output exists only as a dry run:

  • dry-run, preview, read-only, or orient: do not write; print the exact seed entries that would be written
  • refresh: inspect existing seed threads and write additive refresh entries instead of trying to replace prior context
  • --update-agent-context (or update-agent-context): after seeds are written and Step 5.5 verifies the onboarding tag landed, chain into the update-agent-context skill (Phase 1) so CLAUDE.md / AGENTS.md reflect the freshly-seeded threads. Existing files are backed up first (see Step 5.6). No-op in dry-run mode and when Step 5.5 has unresolved tag failures. Without the flag, Step 6 prints the equivalent command as a recommendation instead.
  • research pre-pass (default-on): before deep-history and the seeds, run the Step 2.3 research pre-pass — harvest the subject repo's external references (papers, source links) from the README + docs into an onboarding-biblio thread, and when a principal source paper exists, fetch + parse its bibliography and pull the salient secondary references in too. Auto-skips the network steps under no-biblio / no-github / local-only / offline and in dry-run (prints planned entries, writes nothing); the offline harvest of explicit README links still runs. Detail in references/research-prepass.md.
  • deep-history (opt-in): after the research pre-pass, run the Step 2.4 deep-history / PR-reasoning layer — mine PR history for abandoned/superseded approaches and write the history-* threads. Off by default (forge/PR mining is expensive); honors dry-run (prints planned findings, writes nothing). Detail in references/deep-history.md.
  • role hints (implementer, planner, critic, tester, pm, scribe) shift the recommended entry path and risk emphasis
  • any other text is extra prioritization context

Do not produce a standalone read-only summary as the final artifact. A summary is not repository memory. The useful artifact is a small set of typed, sourced Watercooler entries.

Interpretation guidance, anti-laundering rules, and provenance standards live in references/thesis.md. Load it when claims are uncertain or inferred.


Read the full file on GitHub · 868 lines

Files

What ships with it

5 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.

  1. 10d ago First seen · 868 lines · 79 tokens per session scan A 50c84ec5ae05

Subscribe to this mod's changes

watercooler-onboarding is a skill published in the GitHub repository mostlyharmless-ai/watercooler (13 stars, last pushed 1mo ago), licensed Apache-2.0. It adds 79 tokens to every session and 12,069 once invoked, about $0.0004 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.