Claude Skills by phenobarbital

Automate browser interactions (navigation, form filling, screenshots, data extraction, web-app testing) using a persistent async Playwright session. Use when the user needs to navigate websites, interact with web pages, fill forms, take screenshots, test web applications, or extract information from pages. Activate on tasks mentioning "browser automation", "fill a form", "scrape", "screenshot a page", "test the web app", or "click through a site".

Render deterministic budget-variance facts (the narrative_facts transformer output) as short executive-summary prose. Quote only figures present in the facts; never invent a number.

Reviews code changes and provides feedback, code style and best practices.

Create and refine specification files for Codex-driven implementation through CodeToolkit.

Transform data into compelling narratives using visualization, context, and persuasive structure. Use when presenting analytics to stakeholders, creating data reports, or building executive presentations.

Validates SQL schema files for compliance with internal safety and naming policies.

Document a Python module and its classes using Google style

Formats git commit messages according to Conventional Commits specification. Use this when the user asks to commit changes or write a commit message.

Converts JSON data snippets into Python Pydantic data models.

Adds the standard open-source license header to new source files. Use involves creating new code files that require copyright attribution.

Recursive prompt improvement with quality-driven iteration for Claude Code. Use when you need systematic, measurable improvement of LLM outputs through complexity routing, context extraction, and iterative refinement.

Determine the optimal meta-prompting strategy for any task by analyzing complexity factors and routing to the appropriate approach.

Score LLM output quality (0.0-1.0) against task requirements to determine if iteration is needed or solution is complete.

Analyze LLM outputs to extract patterns, constraints, and success indicators that can improve subsequent iterations.

Recursively improve LLM outputs through quality-driven iteration with automatic complexity routing, context extraction, and quality assessment.

Scaffolds a new custom Tool class for the ai-parrot Tool platform.

Generate production-ready Dockerfiles with multi-stage builds, security best practices, and optimization. Use when containerizing Python applications for production deployment.

Apply Python tooling standards including uv package management, pytest testing, ruff/basedpyright code quality, one-line docstrings, and self-documenting code practices. Use this skill when working with Python backend code, managing dependencies, running tests, or ensuring code quality. Apply when installing packages, writing tests, formatting code, type checking, adding docstrings, organizing imports, or deciding whether to create new files vs. extending existing ones. Use for any Python dev...

Implement or modify a Rust function exposed to Python via PyO3 in eo-processor. Use when adding new compute kernels, fixing numerical/shape bugs in the Rust core, or wiring Rust functions into the Python module safely and consistently.

Use ONLY after the user explicitly approves and requests a PR (keywords: open PR, create pull request, ready for main, approved, ship it). Find the active worktree, verify it is clean, push the branch, create a PR to main, and remove the worktree.

Use when the user asks to implement a feature in an isolated git worktree / separate branch (keywords: worktree, separate folder, isolated branch, new feature branch). Create a new branch + worktree from main, store the active worktree info, and do ALL work inside the worktree. Do NOT open a PR yet.

Use to inspect the current git worktree setup, show the active/likely-active worktree, verify cleanliness, and remind the agent to work only inside the worktree (keywords: worktree status, which worktree, where am I, sanity check, active branch).

Recursive prompt improvement with quality-driven iteration for Claude Code. Use when you need systematic, measurable improvement of LLM outputs through complexity routing, context extraction, and iterative refinement.

Determine the optimal meta-prompting strategy for any task by analyzing complexity factors and routing to the appropriate approach.

Score LLM output quality (0.0-1.0) against task requirements to determine if iteration is needed or solution is complete.

Analyze LLM outputs to extract patterns, constraints, and success indicators that can improve subsequent iterations.

Recursively improve LLM outputs through quality-driven iteration with automatic complexity routing, context extraction, and quality assessment.

How to install a new Odoo module on the running instance. Use this skill when the user asks to install, enable, or activate an Odoo add-on or module. Covers both shell-based installation via odoo-bin and UI-based installation via the Apps menu.

How to answer "how do I do X in Odoo?" questions. Use this skill whenever the user asks a how-to or operational question about Odoo (e.g. "how do I create an invoice?", "how do I install a module?"). Responds with an ordered list of concrete, version-aware steps grounded in the PageIndex documentation.

Svelte 5 + SvelteKit development with structural composition patterns. Use when building dashboards, widget systems, or component libraries where structural consistency matters. Emphasizes base components with behavior composition (strategies), OOP for state machines, and Svelte 5 runes. Triggers on Svelte/SvelteKit work requesting widgets, dashboards, reusable component systems, or when user prefers OOP patterns over pure functional composition.

How to scaffold a Python agent from the base-class catalog, or create a simple YAML-defined agent

Conventions for writing knowledge-base files for an existing agent

The frontmatter contract for writing a valid skill file (per-agent or shared catalog)

**Feature**: FEAT-420 — FinanceReporter Tier-2 + Narrative Skill **Spec**: `sdd/specs/finance-reporter-tier2-narrative.spec.md` **Status**: pending **Priority**: high **Estimated effort**: S (< 2h) **Depends-on**: TASK-2186 **Assigned-to**: unassigned ---