Create and refine AI agent skill definitions with precise trigger contracts.
npx skills add https://github.com/mlesk/darkmark-skills --skill dm-pocockify
Please help me install this Agent Skill. Skill: dm-pocockify Source: https://github.com/mlesk/darkmark-skills/tree/main/skills/productivity/dm-pocockify Command: npx skills add https://github.com/mlesk/darkmark-skills --skill dm-pocockify
Agent skills often have vague trigger descriptions, unclear scope, and no verification steps, leading to inconsistent performance when loaded by AI coding agents. This skill eliminates that friction by enforcing a structured, purpose-first workflow to create or refine skills that load reliably and produce predictable, high-quality outputs.
Use the dm-pocockify skill to create a pocockified copy of your existing 'document-summarizer' skill that only loads when the user asks to summarize legal contracts for key risk clauses.
High-intent search queries and answers about installing and using this skill.
Creating AI agent skills that load reliably requires a trigger-first design with precise load boundaries. This skill provides a structured workflow to define explicit trigger contracts so skills activate only for relevant user requests, avoiding context bloat.
Refining vague agent skill definitions involves enforcing phase gates, one-question-at-a-time user input, and a self-review checklist. This workflow sharpens existing skills into versions with clear exit criteria and predictable outputs for coding agents like Claude Code, Codex, and GitHub Copilot.
Reviewing AI agent skills without editing is supported by a dedicated review mode. This mode evaluates existing skill configurations against phase gates and exit criteria to identify gaps and vague trigger boundaries, producing a validation report instead of modifying the original skill.
Coding agent skills load for unrelated requests due to vague trigger descriptions and unclear scope. Applying a trigger-first design with explicit load boundaries ensures the skill only activates when specific user conditions are met, such as reviewing Python backend code for security issues.
Yes, this skill creation workflow applies to building skills for AI coding agents like Claude Code, Codex, and GitHub Copilot. It ensures skills have precise load boundaries and produce predictable outputs across these platforms.
Designing agent skills without phase gates leads to inconsistent performance and unreliable skill configurations. Without built-in validation and exit criteria, skills lack the necessary structure to catch gaps before finalization, resulting in vague and unpredictable outputs.
gabrielfruetCreate agent skills with valid YAML frontmatter and trigger descriptions.
elliottpolkMap .agentic agents, skills, and workflows to GitHub Copilot prompts.
otrebuImprove AI agent prompts using context engineering and signal-to-noise optimization.
giosuetedeschi-specDesign and refine prompts for consistent LLM outputs.
