skills.rest
data-goblin
data-goblinCOMMUNITY
Explore all skills by this author →
Website𝕏@kurtbuhlerGitHub

improve-my-agent-setup

Audit and improve an agentic-development setup across skills, memory, tools, permissions, and secrets hygiene.

887|131|Updated Jan 15, 2026
Search Tags:#workflow optimization#agentic development#setup audit#configuration review#secrets hygiene#Claude code#Memory Management
One-click install
npx skills add https://github.com/data-goblin/power-bi-agentic-development --skill improve-my-agent-setup
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: improve-my-agent-setup
Source: https://github.com/data-goblin/power-bi-agentic-development/tree/main/plugins/goblin-mode/skills/improve-my-agent-setup
Command: npx skills add https://github.com/data-goblin/power-bi-agentic-development --skill improve-my-agent-setup

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

Agentic-development setups decay over time: unused skills pile up, memory files bloat past their token budget, redundant tools overlap, plaintext secrets leak into git, and safety rules exist only as unenforced sentences. This Skill performs a structured, setup-wide health check that finds the two or three changes that would most improve how you work with AI agents, then applies them with your consent.

Core Features & Use Cases

  • Deterministic setup scan: A stdlib-only Python script gathers facts about installed skills, memory token budgets, settings, hooks, MCP servers, git state, transcripts, other agentic clients (Claude Desktop, Cursor, VS Code, Zed), network exposure, and candidate secrets, without ever writing secret values to disk.
  • Four audit modes: shallow, deep (default), ultra, and yolo modes control scan breadth, interview cadence, and tone, from a quick glance to a machine-wide demolition-and-rebuild review that mines session transcripts.
  • Rubric-driven scoring and fixes: Twenty-plus themes (skills ownership, memory, permission posture, isolation, supply-chain defense, secrets hygiene, and more) are judged against a documented rubric, reported to a dated scratchpad file, and paired with a fix catalog of concrete commands.
  • Use Case: Run a weekly deep audit to catch drift early, such as memory creeping past 15k tokens, a .env file that slipped into git, or a redundant MCP server duplicating a CLI, then apply the ranked fixes one at a time.

Quick Start

Ask the agent to run a deep audit of your agent setup, for example by saying: audit my agentic development setup in deep mode and show me the top fixes.

Frequently Asked Questions about improve-my-agent-setup

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit my Claude Code agent setup?▼

Invoke the skill with an optional mode argument such as deep, ultra, or yolo. It runs a scan script that gathers facts about skills, memory, settings, tools, git, and secrets, then interviews you, scores each theme against a rubric, and writes a report to .claude/scratchpads/.

What is the difference between shallow, deep, ultra, and yolo audit modes?▼

Shallow is a quick glance at basics like tools and memory length. Deep is the default full checkup at user level. Ultra goes machine-wide across all projects with continuous interviewing. Yolo additionally mines session transcripts for error and usage patterns and assumes a rebuild.

Does the setup scan expose my secrets or API keys?▼

No. The scan script is built so secret values never reach disk or output. It reports only a file path, line number, rule name, and match length for candidate secrets, and environment variables are reported by key name only.

Can the audit check other tools like Cursor or VS Code?▼

Yes. The scan detects Claude Desktop, VS Code, Cursor, and Zed configurations, including their MCP servers and rule files, so it can flag redundant capability wired up across multiple agentic clients.

Will the audit change my configuration automatically?▼

No. The audit is read-only by design. It presents findings and a ranked fix list first, then applies each fix only after you confirm, and it never routes around hooks or blocked actions.

When should I not use the yolo audit mode?▼

Avoid yolo for routine checkups because it reads transcript content, costs significant time and tokens, and assumes a demolition-and-rebuild. Use deep for regular reviews and reserve yolo for when your setup is genuinely broken and you want maximal scrutiny.

Related Skills

Explore More Software Engineering Skills→
rickyanandarickyananda
1

hermes-agent

Configure and troubleshoot Hermes Agent installation, runtime settings, and feature access.

Community
Advanced
arbazkhan971arbazkhan971
26

secrets

Detect exposed secrets and rotate credentials into secret managers.

Community
Advanced
joshka0joshka0
3

skill-developer

Author Claude Code skills with SKILL.md frontmatter and trigger rules.

Community
Advanced
Motion-CreativeMotion-Creative
3

creative-deep-dive

Orchestrates mechanism-level creative diagnostics to explain ad performance and recommend next steps.

Official
Advanced