claude-skills

Detect AI writing patterns in text. Use when: (1) need to check if text sounds AI-generated, (2) reviewing content before publication, (3) analyzing writing for Wikipedia-style AI markers, (4) generating report-only analysis without changes.

Use when: (1) User asks to "review for AI writing" or "check for LLM signs" (2) User wants to "detect AI-generated text" or "humanize this document" (3) User wants to identify and fix signs of AI/LLM-generated writing (4) User mentions Wikipedia AI writing guidelines

Fix AI writing patterns in text. Use when: (1) ai-writing-analyzer found issues to fix, (2) need to humanize AI-generated content, (3) preparing content for publication, (4) iteratively refining text until clean.

Humanize a document by removing AI-writing patterns via a multi-agent workflow (analyze → revise → adversarial review → loop-until-clean). Use when the user asks to "humanize", "de-AI", "remove AI tells", or "review for AI writing" on a file.

Fix for Claude Code slash command frontmatter that fails to parse when `argument-hint:` is given square-bracketed placeholder values. Use when: (1) authoring `commands/<name>.md` files in a CC plugin and the frontmatter shows `argument-hint: [working-dir] [team-name]` or similar multi-token bracketed values; (2) a frontmatter-validation test (e.g. `yaml.safe_load(block)` over the SKILL/command frontmatter) raises `ScannerError` or `ConstructorError` complaining about flow sequences or "expected a mapping"; (3) the slash command silently doesn't appear in `/help` after install, or a plugin-marketplace test asserting the `description` field is present trips because the whole frontmatter parse blew up. The fix is to QUOTE the value as a single string. Covers the same trap for `commands/<name>.md` directly authored in CC format AND APM-managed marketplaces shipping commands via APM.

Organize GitHub repository issues with proper milestones, native relationships, and clean labeling structure. Use when grooming issues, setting up milestones, establishing dependencies, or cleaning up issue tracking in a GitHub repository.

Enable Claude Code to delegate specific tasks to Amazon Q CLI agents running in the background with full tool permissions. Use when bridging to Q's specialized integrations such as Quip, ticketing systems, or diagram generation that Claude doesn't have direct access to.

Transfer markdown documents to Quip with proper formatting validation, handling common issues with lists, tables, and diagrams that occur during bulk uploads. Use when writing or syncing markdown content to Quip documents with section-by-section transfer for reliability.

Generate or improve README files through iterative analysis and refinement. Analyzes project codebase, creates or enhances README content following best practices, evaluates quality, and applies improvements until a quality threshold is met or maximum iterations reached.

Generate comprehensive software effort estimation reports that analyze codebase complexity, calculate traditional development effort across five estimation models, and quantify productivity gains from LLM-assisted development. Use when estimating project cost, valuing a codebase, or comparing AI-assisted vs traditional development effort.

Use when user says "create a tdd plan", "tdd loop", "iterative plan", "ralph-loop plan", "create plan for ralph". Generates a TDD-based plan formatted for ralph-loop execution with concurrent tasks, pr-review validation, and optional deploy/integration test steps.

Intelligently route tasks to the most appropriate Amazon Q CLI agent based on task type, with support for explicit agent selection and special workflow patterns. Use when delegating tasks to Q agents or triggering the quip-upload workflow for file uploads to Quip.

Extract reusable skills from Claude Code sessions. Triggers: /claudeception command, "save this as a skill", "what did we learn?". Creates SKILL.md files for non-obvious discoveries, workarounds, and debugging insights.

Fix for Task tool subagents with subagent_type=Bash being unable to write files when PreToolUse hooks block write operations. Use when: (1) dispatching parallel Bash agents to implement code in worktrees, (2) agents report "Permission to use Bash has been auto-denied", (3) agents can read files but all write/touch/cp/dd commands fail. Solution: write files from the parent session using Write/Edit tools, then optionally dispatch Bash agents only for validation (lint, test, typecheck).

Fix AWS CDK bundling failures caused by multiple package lock files in the same project. Use when: (1) CDK synth/deploy fails with "ValidationError: Multiple package lock files found: pnpm-lock.yaml, package-lock.json", (2) NodejsFunction or other bundled Lambda fails during cdk synth, (3) CI pipeline fails on test stage with CDK bundling errors. Covers determining which lock file to keep based on project's actual package manager.

Fix disk space exhaustion caused by AWS CDK temp folder accumulation on macOS. Use when: (1) Disk is full or nearly full with no obvious cause, (2) Running CDK synth, deploy, or tests (especially in TDD loops), (3) /private/var/folders is consuming tens of gigabytes, (4) cdk.out folders accumulating in temp directory, (5) Standard cache cleanup (npm, go, homebrew) doesn't recover space, (6) Running parallel subagents with CDK operations in Claude Code. Covers diagnosis, cleanup, and prevention for CDK projects.

Fix extremely slow AWS CDK tests (30+ minutes for a single test file). Use when: (1) CDK tests take 30+ minutes to run, (2) Each test creates `new cdk.App()` in beforeEach, (3) Tests trigger Docker image builds repeatedly, (4) Jest shows tests running but no progress for minutes, (5) Disk fills up during test runs (see also: cdk-temp-folder-disk-bloat). Solution: Use beforeAll to share CDK synthesis across tests.

How to discover and add Claude Code plugin marketplaces. Use when: (1) user asks about claudemarketplaces.com or similar discovery sites, (2) user wants to add a plugin marketplace, (3) confusion between marketplace directories and actual marketplaces, (4) user runs /plugin marketplace commands. Covers the distinction between discovery directories and installable marketplaces.

Analyze Claude Code sessions to extract configuration improvements. Use when: (1) Want to optimize Claude Code setup (skills, permissions, CLAUDE.md), (2) Need to find what tools should be pre-granted (repeated permission requests), (3) Looking for skills to create (repeated investigation patterns), (4) Finding rules to add to CLAUDE.md (user corrections indicate missing guidance), (5) Diagnosing why sessions went poorly (compactions, interruptions). The goal is actionable improvements, not just metrics.

Fix CloudFormation deployment failures caused by KMS alias collisions. Use when: (1) CDK deploy fails with "AWS::EarlyValidation::ResourceExistenceCheck" error, (2) Multiple CDK stacks create KMS keys with similar aliases, (3) Deployment fails before changeset execution due to resource existence check. Covers diagnosing KMS alias conflicts and resolving them in CDK stacks.

Learned from user correction: Correction: Avoid: 'build locally'

Learned from user correction: Correction: Avoid: 'cancel!'

Learned from user correction: Correction: Avoid: 'cite genai-powered sdlc guide as a source'

Learned from user correction: Correction: Avoid: 'delete 670'

Learned from user correction: Correction: Avoid: 'delete recordings'

Learned from user correction: Correction: Avoid: 'delte the vectors!!!'

Learned from user correction: Correction: Avoid: 'deploy yet until we have a better picture of what happened in v2' | Expected behavior: 'match v2's should be the same as v2's'

Learned from user correction: Correction: Avoid: 'do a separte plugin'

Learned from user correction: Correction: Avoid: 'do it again'

Learned from user correction: Correction: Avoid: 'do it till i tell you &'

Learned from user correction: Correction: Avoid: 'do that'

Learned from user correction: Correction: Avoid: 'do the loop manually'

Learned from user correction: Correction: Avoid: 'forget it'

Learned from user correction: Correction: Avoid: 'have to wait for 670 to merge to retrigger'

Learned from user correction: Correction: Avoid: 'investigate. create a task and spawn an agent to investigate'

Learned from user correction: Correction: Avoid: 'map clearly to any specific sdlc phase: - `normalize-structure-after-exploration' | Expected behavior: 'be tagged with which sdlc phases based on my review of all 73 files (notes 293-365)'

Learned from user correction: Correction: Avoid: 'mention any of the gitlab common repos we have with toosl'

Learned from user correction: Correction: Avoid: 'mention github anywhere'

Learned from user correction: Correction: Avoid: 'miss any knowledge' | Expected behavior: 'include revieweing existing hubs and their attached links and performing a gathering of the relevant notes for the given section before startin on the section to make sure we dont miss any knowledge'

Learned from user correction: Correction: Avoid: 'misss any info'