fb-claude-skills

Structured, user-invoked standing prompts for long runs on any project: /design-scoreboard chooses what to climb and what must not get worse, /improve and /optimize climb it against the repo's AGENTS.md sections, and /trim-agents-md keeps that hub short. Each keeps state in files and reports what needs the owner first

4 skills · 2 agents · hooks

MECE decomposition for breaking work into mutually exclusive, collectively exhaustive parts. A skill you invoke when decomposing something -- it ships no SessionStart hook, because the method is needed at a decision point the model can recognise, not before the first action of every session.

5 skills · 1 MCP server

The audit family: evidence-grounded postmortems of finished work with a browsable index, a test-suite audit (does each green test still mean anything), a controls audit (census and live-fire over hooks, validators, and reminders), and the adversarial-verify primitive with its control-builder agent that the audits dispatch to

5 skills · 1 agent

Skills for how the agent writes and explains: an American plain-language house style, voice matching, replacing prose with compact visuals, and re-pitching a message that did not land

4 skills · 1 command

Hand a task to Gemini and get a structured answer back -- video and audio analysis, perceptual image comparison, or any ad-hoc question with model, thinking level, and schema set per call. Explicit, auditable, one run directory per call.

1 skill · 2 commands · hooks

On-demand references for what a repo cannot state by being read: the house Python pinning policy and the two mistakes behind most Pydantic/Pyright diagnostic walls; documentation conventions (last-updated dates, where unshared notes and session logs live, how dependency changes are recorded, no decorative counts in prose); and a uv and bun dependency CVE audit, transitives included. No hooks or ambient injection: conventions belong in each repo's own always-loaded files.

3 skills

Tools for analyzing, polishing, and managing Claude Code plugins

1 skill · 2 agents

Keeps personal identifiers out of repo content: every path written into the repo must be relative to the repo root, and the git user.name full name never appears. PreToolUse hooks rewrite in-repo absolute paths to relative and block external paths or full-name leaks across file edits and git/gh commands; pre-commit and commit-msg git hooks hard-block at commit.

1 skill · hooks

Audit prose as untrusted claims -- a diff's added lines, a standing doc such as a README or AGENTS.md checked against today's code, or a summary about to be sent. Every count, status, capability, pointer and number is re-derived by running a command, never by reading; --fix applies only the corrections the evidence settles. Reports claims beside their deriving commands and states its own scope.

1 skill

Remove a unit from a repo without leaving references behind. Sweeps tracked content before you delete, sorts every hit into what must change, what must stay as history, and what reaches users, and names the cascade a removal actually triggers.

1 skill

Dimensional modeling for analytical schemas: grain-first design, append-only facts, SCD Type 2 dimensions, conformed dimensions instead of fact-to-fact joins. A skill you invoke when designing a schema -- it ships no SessionStart hook, because the principles are needed at a decision point the model can recognise, not before the first action of every session.

1 skill

A relentless design interview: work the problem as a tree, ask every question whose prerequisites are settled in one round, and treat fact-finding as the agent's job rather than the user's

1 skill

Wire an application to heylook (heylookitsanllm) as a local inference provider: the Anthropic Messages-conformant /v1/messages wire, runtime capability discovery, porting off the removed OpenAI route, and the deliberate differences from Anthropic's spec

1 skill

JSON query tool selection and syntax guidance — jg (jsongrep) for fast extraction, jq for transformations

1 skill

Per-project down-tier model delegation: installs a standalone .claude/rules/ file telling Claude to route well-specified data and coding tasks to a cheaper model in a subagent, keeping judgment-heavy work in the main loop, with optional fast-executor / task-coder agent definitions. Installation is paused pending measurement and the skill is user-invoked only; removal is unaffected.

1 skill

Write an implementation plan as one interactive HTML page: a tree of claims, each proved by a mockup, state machine, call stack, schema or code, with the decisions you need to make placed where they matter. Use /plan-page <what to build>.

1 skill

SessionStart hook that drops a personal, locally-git-ignored pyrightconfig.json into any Python project lacking a Pyright config -- points Pyright at the uv .venv so imports resolve and silences the reportMissingImports flood. Retracts that file if the project later declares its own [tool.pyright], since Pyright ranks pyrightconfig.json above pyproject.toml and ours would otherwise outrank it forever. Never touches a config it did not write byte-for-byte; no-op outside Python projects.

hooks

PostToolUse hook that runs Ruff on each edited .py/.pyi file and reports findings to Claude. Claude Code starts only one language server per file extension, so `ruff server` can never run alongside pyright-lsp; this reconstructs the standard Pyright-plus-Ruff split from outside the LSP layer. Read-only: never writes ruff config, never syncs the environment, silent on clean files.

hooks

Pre-share scanner: wraps simonw/scan-for-secrets for literal matches and composes a ripgrep regex pass to catch leaked secrets and privacy-sensitive paths

1 skill

Interactive MCP App skill dashboard: quality checks, token budgets, and repo hygiene

1 skill

Version bumps across every file that carries a plugin's version: plugin.json, the marketplace entry and a shipped pyproject.toml (/sync-versions). The checks themselves are the separate skill-maintain CLI.

1 skill