Sandeeprdy1729

A project completion skill that transforms nearly finished projects into complete ones by identifying specific gaps and the next action needed to unblock shipping.

This skill synthesizes large documents and codebases (50k–200k+ tokens), providing perfect recall for synthesis, cross-referencing, and insight extraction. It's a precision instrument for finding patterns, contradictions, and dependencies, not a summarizer.

A self-critique and Constitutional AI reasoning skill. It enables Claude to evaluate and revise its own outputs against defined principles, reducing hallucination and over-confidence by forcing it to argue against its answers.

This skill provides automated PR review and code quality analysis. It activates for tasks like reviewing pull requests, auditing code, or scanning for issues, fanning out parallel sub-agents for various review lenses to synthesize a structured report.

This constraint-based problem-solving skill finds elegant solutions within heavily restricted design spaces, such as budget caps or regulatory requirements. It treats constraints as design material, identifying violations early and finding solutions that satisfy all hard constraints.

MCP tool deduplication and unified abstraction layer for multi-server setups. Activates when multiple MCP servers are installed to reduce tool noise, detect overlapping tools, guide tool selection, or create routing rules to prevent token waste on near-identical tools. Handles semantic duplicate detection across MCP servers.

This adversarial red-teaming skill attacks code, systems, strategies, and plans to uncover security holes, logical flaws, incorrect assumptions, and unconsidered risks. It activates when the user wants their work rigorously challenged.

A skill for writing voice authentication and AI-smell elimination. It activates when the user wants AI-generated writing to sound specifically like them, analyzing a writing sample to replicate their unique style.

This rapid iterative refinement skill applies 4-6 structured improvement cycles to any output, from writing to code. It ensures coherence and prevents regression or drift from the original intent, with each cycle targeting a specific dimension until a quality threshold is met.

This persistent context management skill eliminates the need for users to repeatedly explain their role, project, or constraints. It activates when Claude's answers lack context, building a structured template to inject relevant information per question type.

A self-healing skill for generating and maintaining CLAUDE.md files. It activates to create, update, or repair CLAUDE.mds for a project, or after errors, by scanning the codebase to generate a concise root CLAUDE.md.

This meta-skill and skill manager activates as the primary entry point, processing incoming queries to select, load, and execute relevant skills and tools, then unloads them to keep the context clean.

Target voice and style mimicry skill. It copies any external writing style, such as a brand or public figure, from provided samples.

Skill trigger debugging and description auditor. Activates when a skill is not triggering, triggering on the wrong query, too eagerly, or conflicting with another skill. It analyzes the skill's description, tests trigger coverage, detects negative triggers, flags description overlaps, and produces a rewritten description with a test suite.

A decision-forcing skill that eliminates hedging and commits to a single answer. Use it when the user wants Claude to choose between options, recommend an approach, or make a comparative judgment, seeking a definitive answer rather than a pros-and-cons list.

An adversarial pre-mortem skill that actively hunts for fatal flaws in plans, proposals, and strategies. It refuses to be supportive, activating when the user wants their plan stress-tested or attacked before committing, surfacing the 1-3 most vulnerable assumptions.

A silent failure detection skill that identifies when Claude is confidently wrong, surfacing hidden assumptions, overconfident claims, and hallucinations. It uses specific questioning techniques to reveal actual uncertainty, transforming overconfidence into a detectable signal.

This skill identifies root causes and rephrases questions, answering the underlying intent when a user's query stems from an incorrect mental model. It performs a 3-step check (user's goal, question's relevance, faster path) to surface reframes and address XY problems.

This skill enforces strict structured output, ensuring Claude produces data in exact, machine-readable formats like JSON, YAML, or CSV, with zero deviation or schema violations. It validates the output against the schema before delivery, activating when parsed or imported output is required.

This multi-step prompt chaining skill uses the output of each step as the input for the next, with explicit rules and constraints. It activates for complex pipelines of dependent Claude calls, ensuring downstream steps are constrained by upstream outputs to prevent hallucination or drift.

This skill combines knowledge from different fields to generate novel, practical insights, applying mechanisms from one domain to solve problems in another. It activates when unconventional approaches are sought, or when users are stuck in standard thinking and wish to borrow solutions.