aiskillstore/marketplace / skills/aaronabuusama/prompting
prompting
VerifiedPrompt engineering standards and context engineering principles for AI agents based on Anthropic best practices. Covers clarity, structure, progressive discovery, and optimization for signal-to-noise ratio.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 2.6 KB |
| skill-report.json | config | 18.5 KB |
| workflows/create-prompt.md | markdown | 1.1 KB |
| workflows/optimize-prompt.md | markdown | 1.4 KB |
skill-report.json
View on GitHub{
"schema_version": "2.0",
"meta": {
"generated_at": "2026-07-23T10:03:11.840Z",
"slug": "aaronabuusama-prompting",
"source_url": "https://github.com/AaronAbuUsama/alpha-arena/tree/develop/.claude/skills/prompting",
"source_ref": "develop",
"model": "claude",
"analysis_version": "3.0.0",
"source_type": "community",
"content_hash": "ceaa98f4ff0ee951fc4410de8a7832fb841a8c1a7d4bbae8bf2efbb5fae9b950",
"tree_hash": "f0bff9d6be56d4e1be4749a0377f01c2f9bcab76eacdf3c220606ee47cea4bbd",
"provenance": {
"model_effective": "claude",
"fallback_chain": [
{
"agent": "codex",
"outcome": "failed",
"error": "2026-07-23T10:03:11.950596Z ERROR codex_models_manager::manager: failed to refresh available models: Missing environment variable: `HELM_API_KEY`.\n2026-07-23T10:03:11.978187Z ERROR codex_models_manage…"
},
{
"agent": "claude",
"outcome": "success"
}
],
"analysis_version": "3.0.0"
},
"upstream_version_raw": null,
"upstream_version_normalized": null,
"upstream_version_source": "none",
"upstream_version_status": "missing",
"upstream_commit_sha": "5387935c03c9084f473472bc526b5237c8c0a027"
},
"skill": {
"name": "prompting",
"description": "Prompt engineering and context engineering standards for AI agents, based on Anthropic best practices. Provides guidance on clarity, semantic structure, progressive information discovery, and signal-to-noise optimization, plus templates and workflows for creating and improving prompts.",
"summary": "Prompt and context engineering standards for building high-signal AI agent prompts.",
"icon": "📦",
"author": "AaronAbuUsama",
"license": "MIT",
"category": "productivity",
"tags": [
"prompt-engineering",
"context-engineering",
"ai-agents",
"documentation",
"best-practices"
],
"supported_tools": [
"claude",
"codex",
"claude-code"
],
"risk_factors": [
"external_commands"
]
},
"security_audit": {
"risk_level": "safe",
"is_blocked": false,
"safe_to_publish": true,
"analysis_status": "ok",
"summary": "This skill is documentation-only: Markdown files describing prompt- and context-engineering standards, with two template workflows. All ten static findings are false positives. The 'Ruby/shell backtick execution' matches are Markdown inline code spans (identifiers like `calculate_tax` and `tool_name(params)`) used as prose examples, and the 'system reconnaissance' matches are ordinary English sentences. No scripts, network access, filesystem writes, environment access, or prompt-injection content is present.",
"remediation": [],
"risk_factor_evidence": [
{
"factor": "external_commands",
"evidence": [
{
"file": "CLAUDE.md",
"line_start": 98,
"line_end": 99
},
{
"file": "CLAUDE.md",
"line_start": 161,
"line_end": 163
},
{
"file": "CLAUDE.md",
"line_start": 232,
"line_end": 233
},
{
"file": "CLAUDE.md",
"line_start": 233,
"line_end": 246
},
{
"file": "CLAUDE.md",
"line_start": 263,
"line_end": 264
},
{
"file": "CLAUDE.md",
"line_start": 264,
"line_end": 267
},
{
"file": "SKILL.md",
"line_start": 89,
"line_end": 89
}
]
}
],
"static_findings": [
{
"id": "external_commands:CLAUDE.md:98:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "CLAUDE.md",
"line_start": 98,
"line_end": 99,
"snippet": "Use the `calculate_tax` tool with amount and jurisdiction parameters."
},
{
"id": "external_commands:CLAUDE.md:161:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "CLAUDE.md",
"line_start": 161,
"line_end": 163,
"snippet": "Use `get_product(sku)` to retrieve product details when needed."
},
{
"id": "external_commands:CLAUDE.md:232:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "CLAUDE.md",
"line_start": 232,
"line_end": 233,
"snippet": "- `tool_name(params)` - [Brief description]"
},
{
"id": "external_commands:CLAUDE.md:233:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "CLAUDE.md",
"line_start": 233,
"line_end": 246,
"snippet": "- `tool_name2(params)` - [Brief description]"
},
{
"id": "external_commands:CLAUDE.md:263:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "CLAUDE.md",
"line_start": 263,
"line_end": 264,
"snippet": "- `param1` (required): [Description]"
},
{
"id": "external_commands:CLAUDE.md:264:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "CLAUDE.md",
"line_start": 264,
"line_end": 267,
"snippet": "- `param2` (optional): [Description]"
},
{
"id": "blocker:CLAUDE.md:58:system-reconnaissance",
"category": "blocker",
"severity": "low",
"pattern": "System reconnaissance",
"file": "CLAUDE.md",
"line_start": 58,
"line_end": 58,
"snippet": "- Avoid historical details unless critical"
},
{
"id": "blocker:CLAUDE.md:91:system-reconnaissance",
"category": "blocker",
"severity": "low",
"pattern": "System reconnaissance",
"file": "CLAUDE.md",
"line_start": 91,
"line_end": 91,
"snippet": "You should always make sure to validate the user's input before you process it because invalid input"
},
{
"id": "blocker:CLAUDE.md:294:system-reconnaissance",
"category": "blocker",
"severity": "low",
"pattern": "System reconnaissance",
"file": "CLAUDE.md",
"line_start": 294,
"line_end": 296,
"snippet": "## Anti-Patterns to Avoid"
},
{
"id": "external_commands:SKILL.md:89:ruby-shell-backtick-execution",
"category": "external_commands",
"severity": "medium",
"pattern": "Ruby/shell backtick execution",
"file": "SKILL.md",
"line_start": 89,
"line_end": 89,
"snippet": "For full standards: `read ${PAI_DIR}/skills/prompting/CLAUDE.md`"
}
],
"finding_verdicts": [
{
"id": "external_commands:CLAUDE.md:98:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.97,
"reason": "The backticks wrap the identifier `calculate_tax` as a Markdown inline code span inside a prose example sentence. There is no shell or Ruby backtick command substitution; the file is documentation and is never executed."
},
{
"id": "external_commands:CLAUDE.md:161:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.97,
"reason": "`get_product(sku)` is a Markdown inline code span illustrating a just-in-time data-loading pattern in prose. No command execution occurs; it is descriptive documentation."
},
{
"id": "external_commands:CLAUDE.md:232:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.97,
"reason": "`tool_name(params)` is a placeholder identifier in an agent-context Markdown template. It is inline code formatting in documentation, not shell/Ruby backtick execution."
},
{
"id": "external_commands:CLAUDE.md:233:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.97,
"reason": "`tool_name2(params)` is another placeholder identifier in the same documentation template. Markdown inline code, not an executed command."
},
{
"id": "external_commands:CLAUDE.md:263:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.97,
"reason": "`param1` is a Markdown inline code span in a command-template documentation section. No backtick command substitution is present."
},
{
"id": "external_commands:CLAUDE.md:264:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.97,
"reason": "`param2` is a Markdown inline code span in the same documentation template. Inline code formatting, not shell execution."
},
{
"id": "blocker:CLAUDE.md:58:system-reconnaissance",
"verdict": "false_positive",
"confidence": 0.98,
"reason": "The line 'Avoid historical details unless critical' is ordinary English guidance about prompt content. It contains no system-reconnaissance activity or intent."
},
{
"id": "blocker:CLAUDE.md:91:system-reconnaissance",
"verdict": "false_positive",
"confidence": 0.98,
"reason": "This is a verbose-prose example illustrating writing style ('validate the user's input'). It is a documentation example, not a probe of the host system."
},
{
"id": "blocker:CLAUDE.md:294:system-reconnaissance",
"verdict": "false_positive",
"confidence": 0.98,
"reason": "'## Anti-Patterns to Avoid' is a Markdown section header in the prompt-engineering guide. No reconnaissance of the environment is implied."
},
{
"id": "external_commands:SKILL.md:89:ruby-shell-backtick-execution",
"verdict": "false_positive",
"confidence": 0.9,
"reason": "The inline code `read ${PAI_DIR}/skills/prompting/CLAUDE.md` is a documentation pointer telling the reader where the full standards live, referencing a file inside the skill itself. It is not executed as a backtick command and reads only the skill's own bundled documentation."
}
],
"semantic_findings": [],
"critical_findings": [],
"high_findings": [],
"medium_findings": [],
"low_findings": [],
"dangerous_patterns": [],
"files_scanned": 4,
"total_lines": 536,
"audit_model": "claude",
"audited_at": "2026-07-23T10:03:11.840Z",
"agent_auto_install_policy": "allowed",
"manual_install_policy": "allowed"
},
"content": {
"user_title": "Master Prompt and Context Engineering for AI Agents",
"value_statement": "Vague, bloated prompts waste context and produce unreliable AI agent behavior. This skill applies Anthropic context-engineering principles to help you write clear, structured, high-signal prompts that improve accuracy and reduce token cost.",
"seo_keywords": [
"prompt engineering",
"context engineering",
"AI agent prompts",
"Claude",
"Codex",
"Claude Code",
"prompt optimization",
"system prompt design",
"LLM prompting best practices",
"signal-to-noise ratio"
],
"actual_capabilities": [
"Explains context engineering principles: finite attention budget, signal-to-noise optimization, and progressive information discovery.",
"Provides Markdown structure standards with distinct Background, Instructions, Examples, and Constraints sections.",
"Supplies reusable templates for context files, agent definitions, and command definitions.",
"Includes a create-prompt workflow that guides building a new prompt from goal to tested result.",
"Includes an optimize-prompt workflow that reduces verbosity, removes redundancy, and adds structure to existing prompts.",
"Offers a best-practices checklist and a list of anti-patterns to review prompts against."
],
"limitations": [
"Provides written guidance and templates only; it does not run code, call tools, or automatically rewrite prompts.",
"Focuses on prompt and context structure, not on model-specific tuning, evaluation harnesses, or fine-tuning.",
"Does not measure token counts or performance for you; testing and iteration remain manual.",
"Principles are general and may need adaptation for a specific model, domain, or product."
],
"use_cases": [
{
"title": "Design a new system prompt",
"description": "A developer building an AI agent uses the create-prompt workflow and template sections to write a clear, well-structured system prompt from the start.",
"target_user": "AI application developer"
},
{
"title": "Trim a bloated prompt",
"description": "An engineer applies the optimize-prompt workflow to cut redundant text, convert suggestions into direct instructions, and reduce context cost while keeping quality.",
"target_user": "Prompt engineer"
},
{
"title": "Standardize team prompt quality",
"description": "A team lead adopts the best-practices checklist and anti-pattern list as a review standard so all agent prompts across the team follow the same conventions.",
"target_user": "Engineering team lead"
}
],
"prompt_templates": [
{
"title": "Get started with prompt structure",
"prompt": "Using the prompting skill, help me structure a prompt for an assistant that summarizes support tickets. Organize it into Background, Instructions, Examples, and Constraints sections.",
"scenario": "Beginner: first structured prompt"
},
{
"title": "Optimize an existing prompt",
"prompt": "Apply the optimize-prompt workflow to this prompt. Remove redundancy, make instructions direct, and add clear structure. Show the before and after.",
"scenario": "Intermediate: improving a draft"
},
{
"title": "Apply progressive information discovery",
"prompt": "Review this prompt and rewrite it to use just-in-time context loading. Replace inline data dumps with lightweight references that load detail only when needed.",
"scenario": "Advanced: context-cost reduction"
},
{
"title": "Audit a prompt against the checklist",
"prompt": "Evaluate this agent prompt against the prompting skill best-practices checklist and anti-patterns. List each item that passes or fails and give a concrete fix for every failure.",
"scenario": "Advanced: quality review and remediation"
}
],
"output_examples": [
{
"input": "Help me structure a prompt for a code review assistant.",
"output": "The skill returns a prompt organized into four sections: Background (the assistant reviews pull requests for correctness and style), Instructions (imperative steps such as read the diff, flag defects, suggest fixes), Examples (a sample finding), and Constraints (report format, severity levels, and scope limits)."
},
{
"input": "Optimize a verbose 300-word prompt full of hedging language.",
"output": "The skill rewrites the prompt into a shorter version: suggestions become direct instructions, repeated points are merged, paragraphs become bulleted lists, and long inline examples are replaced with references. The result covers the same requirements in far fewer tokens."
}
],
"best_practices": [
"Structure every prompt with clear Markdown sections so each part has one purpose.",
"Write instructions in direct imperative voice and remove hedging words like might, could, and should consider.",
"Load detailed data just in time through references instead of dumping full context up front."
],
"anti_patterns": [
"Filling the prompt with verbose explanations and historical background that add tokens without adding signal.",
"Repeating the same instruction or overlapping tool definitions across multiple places.",
"Loading large data sets or full schemas into the prompt before they are actually needed."
],
"faq": [
{
"question": "What does this skill do?",
"answer": "It provides standards, templates, and workflows for writing and improving prompts for AI agents, based on Anthropic context-engineering principles."
},
{
"question": "Does the skill run any code or commands?",
"answer": "No. It contains only Markdown documentation and templates. It does not execute scripts, access the network, or modify files."
},
{
"question": "Which AI tools can use this skill?",
"answer": "It is written for Claude, Codex, and Claude Code, and its principles apply to prompting for large language models in general."
},
{
"question": "Is this for beginners or experts?",
"answer": "Both. Beginners get clear structure templates and a checklist, while experienced engineers get context-cost reduction and progressive-discovery techniques."
},
{
"question": "What is context engineering?",
"answer": "It is the practice of curating the smallest set of high-signal tokens during inference so the model has exactly what it needs and nothing that wastes its attention budget."
},
{
"question": "Will it automatically rewrite my prompts?",
"answer": "It guides the rewrite through workflows and checklists, but you review and apply the changes. Testing and iteration remain in your control."
}
]
},
"file_structure": [
{
"name": "CLAUDE.md",
"type": "file",
"path": "CLAUDE.md",
"lines": 337
},
{
"name": "SKILL.md",
"type": "file",
"path": "SKILL.md",
"lines": 93
},
{
"name": "workflows",
"type": "dir",
"path": "workflows",
"children": [
{
"name": "create-prompt.md",
"type": "file",
"path": "workflows/create-prompt.md",
"lines": 52
},
{
"name": "optimize-prompt.md",
"type": "file",
"path": "workflows/optimize-prompt.md",
"lines": 54
}
]
}
]
}
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |