robkom-skills
Rob Komaromi's personal Claude Code marketplace.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Check it deeper on your machine
It has changed since its scan on 19 Nov 2026.
As a Claude Code plugin, install it and then ask Claude Code:
/plugin marketplace add keithmackay/mackayi/plugin install scanskillsafety@mackayiDo a deep scan of https://github.com/chipit24/robkom-skills with scanskillsafety
Or from a terminal, clone it and run it:
git clone https://github.com/keithmackay/scanskillsafety-skill ~/.claude/skills/scanskillsafetynode ~/.claude/skills/scanskillsafety/cli.cjs --deep 'https://github.com/chipit24/robkom-skills'
It checks the exact commit: every file, its packages against OSV (api.osv.dev), how it compares with this page's scan (findsafeskills.com receives the repository's address), and, if you agree, Claude reading every file with your own Claude subscription. It never runs the code.
Badge for your README
Markdown
[](/listings/chipit24--robkom-skills)HTML
<a href="/listings/chipit24--robkom-skills"><img src="/api/listings/chipit24--robkom-skills/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
claude plugin marketplace add chipit24/robkom-skills claude plugin install robkom@robkom-skills
Skills are namespaced by the plugin, e.g. `/robkom:code-comments`.
Auto-update is off by default for marketplaces outside of Anthropic's. You can turn it on in `/plugin` → Marketplaces → robkom-skills → Enable auto-update.
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
.claude-plugin/marketplace.json · Manifest · 16 lines
- {
- "name": "robkom-skills",
- "owner": {
- "name": "Rob Komaromi",
- "url": "https://github.com/chipit24"
- },
- "description": "Rob Komaromi's personal Claude Code marketplace.",
- "plugins": [
- {
- "name": "robkom",
- "source": "./plugins/robkom",
- "description": "Rob's general-purpose engineering skills and output styles.",
- "category": "development"
- }
- ]
- }
README.md · README · 108 lines
- # robkom-skills
- My personal Claude Code plugin marketplace, which ships one plugin (`robkom`). I run its skills (and output style) in the Claude Code desktop app, usually with Fable 5.2 or Opus 5.5 at Medium or higher effort level.
- ## Installation
- ```bash
- claude plugin marketplace add chipit24/robkom-skills
- claude plugin install robkom@robkom-skills
- ```
- Skills are namespaced by the plugin, e.g. `/robkom:code-comments`.
- Auto-update is off by default for marketplaces outside of Anthropic's. You can turn it on in `/plugin` → Marketplaces → robkom-skills → Enable auto-update.
- ## Skills
- | Skill | What it does |
- | --- | --- |
- | [code-comments](plugins/robkom/skills/code-comments/SKILL.md) | Makes sure comments explain why, not what. A comment that states a rule becomes a type, assertion, or test instead. |
- | [full-review](plugins/robkom/skills/full-review/SKILL.md) | Reviews the current branch in a new subagent, verifies each finding by running the code, fixes the findings worth fixing in one commit each, and re-reviews, for at most three reviews. It commits its fixes, so it runs only when you type `/robkom:full-review`. |
- | [parallel-worktree-setup](plugins/robkom/skills/parallel-worktree-setup/SKILL.md) | Sets a project up so its app can run from several git worktrees at the same time, each with its own ports, sign-in and data. Adds scripts to provision a worktree, tear it down, and clean up after deleted worktrees. It changes the project, so it runs only when you type `/robkom:parallel-worktree-setup`. |
- ## Output style
- [Clear](plugins/robkom/output-styles/clear.md) leads with the result, like the built-in Concise style, but never cuts the words a reader needs. Replies, PR descriptions, commit messages, docs and comments use full sentences that someone new to the work can follow.
- To use it, set `outputStyle` in `~/.claude/settings.json`:
- ```json
- {
- "outputStyle": "robkom:Clear"
- }
- ```
- The value is case-sensitive. A value that doesn't match exactly gives the Default style, with no error. Claude Code reads style files when it starts, so restart it after you edit the style.
- ## Layout
- ```text
- .claude-plugin/marketplace.json Marketplace catalog
- plugins/robkom/
- .claude-plugin/plugin.json Plugin manifest
- skills/<name>/SKILL.md One folder per skill
- output-styles/<name>.md One file per output style
- scripts/validate.ts Validation run locally and in CI
- .githooks/pre-push Runs the validation before every push
- drafts/ Skills under review (gitignored)
- ```
- ## Development
- On the machine with this clone, register the clone itself instead of GitHub. Skills then load straight from the working tree, and `/reload-plugins` picks up edits without a push:
- ```bash
- claude plugin marketplace add ~/Code/robkom-skills
- claude plugin install robkom@robkom-skills
- ```
- A machine can register the marketplace from only one place, since the GitHub copy and the local clone both use the name `robkom-skills`. Remove one before adding the other, which also uninstalls its plugin.
- To try the plugin in a single session without installing it:
- ```bash
- claude --plugin-dir ./plugins/robkom
- ```
- Install the dependencies with [Bun](https://bun.com). They include a pinned Claude Code:
- ```bash
- bun install
- ```
- CI validates the marketplace, plugin, skills, and output styles on pull requests and pushes to `main`. The checks also confirm that the Skills table lists every skill. To run them locally:
- ```bash
- bun run validate
- ```
- A pre-push hook runs the same checks before every push. Turn it on once per clone:
- ```bash
- git config core.hooksPath .githooks
- ```
- `git push --no-verify` skips it for one push.
- To run the pinned Claude Code, use `bun run claude <args>`.
- ## Skill philosophy
- Every skill in this repo follows these rules:
- - **A skill holds what the model can't know by itself.** That means the author's preferences, their decisions, and the facts of their setup. If a model could have written the whole skill with no input from its author, the skill isn't needed. A skill leaves out anything a capable model would most likely do correctly without being told.
- - **A skill describes the outcome and leaves the method to the model.** It says what must be true when the work is done, and trusts the model to work out how to get there. This keeps the skill useful as models improve. A skill that tells the model how to do the work can get worse, because its instructions may constrain or confuse a model that would have done better without them. A skill that describes the outcome leaves room for a more capable model to do the work better.
- - **A skill informs the model's judgment and doesn't replace it.** An instruction comes with its reason where the reason isn't obvious, so the model can tell where the instruction applies and where it doesn't. A technique that has worked before is offered as an option, for the model to use where it fits.
- ## Adding a skill
- 1. Draft it in `drafts/<name>/SKILL.md`, following the [skill philosophy](#skill-philosophy).
- 2. Once it's reviewed, move the folder to `plugins/robkom/skills/<name>/` and add it to the Skills table.
- 3. Commit and push. The pre-push hook and CI validate it.
- ## Versioning
- Neither manifest sets `version` yet. A marketplace added from a local path ignores `version` anyway and loads the working tree at every session start. An install from GitHub without a `version` uses the commit SHA instead, so every push is an update.
- To pin releases later, add `version` to `plugin.json` and bump it on every release. GitHub installs ignore new commits until it changes.
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.