skills.rest
jrmatherly
jrmatherlyCOMMUNITY
Explore all skills by this author →
WebsiteGitHub

prompt-review

Audit Spacebot system prompts for structural issues, token waste, behavioral drift, and injection risks.

Updated Apr 14, 2026
Search Tags:#audit#prompt-engineering#token-efficiency#system-prompt#prompt-injection#prompt-review#spacebot
One-click install
npx skills add https://github.com/jrmatherly/spacebot --skill prompt-review-jrmatherly
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: prompt-review
Source: https://github.com/jrmatherly/spacebot/tree/main/.claude/skills/prompt-review
Command: npx skills add https://github.com/jrmatherly/spacebot --skill prompt-review-jrmatherly

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Identifies structural issues, behavioral drift, token waste, and prompt-injection risks in a Spacebot agent's live system prompt so operators can fix root causes and reduce incorrect or unsafe model behavior.

Core Features & Use Cases

  • Live Prompt Inspection: Pulls the fully rendered system prompt from the Spacebot inspect API and surfaces composition details and metadata.
  • Layered Analysis: Diagnoses problems by layer (Identity files, channel template, dynamic fragments, knowledge synthesis, working memory, runtime status) and points to the exact source files or templates to edit.
  • Safety & Efficiency Checks: Flags prompt injection vectors, redundant learned memories, excessive token spend, stale facts, and cross-process inconsistencies for actionable remediation.
  • Use Case: Run when the agent exhibits unexpected behavior, produces verbose or contradictory responses, or consumes excessive context tokens.

Quick Start

Ask the agent to audit the live system prompt and produce a prioritized report of critical issues, token efficiency improvements, and exact file-or-template locations to fix them.

Frequently Asked Questions about prompt-review

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

FAQPage Schema
How do I audit a system prompt for token waste and behavioral drift?▼

System prompt injection risks can be identified by inspecting the live rendered prompt and checking dynamic fragments, knowledge synthesis, and working memory layers. This audit flags potential injection vectors and cross-process inconsistencies for actionable security fixes.

How do I find the source template causing incorrect agent behavior?▼

Finding the source template causing incorrect agent behavior involves diagnosing identity files, channel templates, and runtime status fragments. The audit attributes issues to exact files or Jinja template locations so operators can fix root causes directly.

When should I run a prompt review on my agent?▼

A prompt review should be run when an agent exhibits unexpected behavior, produces verbose or contradictory responses, or consumes excessive context tokens. It diagnoses structural issues and behavioral drift across channel, worker, branch, and cortex prompts.

Do I need the Spacebot inspect API to review agent prompts?▼

Yes, reviewing agent prompts requires access to the Spacebot inspect API to pull the fully rendered system prompt. It also needs agent identity files on disk and prompt template sources to attribute and recommend fixes accurately.

Related Skills

Explore More Software Engineering Skills→
CaneproCanepro

sentry

Fetch and inspect Sentry issues and events via a read-only API.

Community
Advanced
jrmatherlyjrmatherly

cluster-ops

Inspect and recover Spacebot deployments in a Talos Kubernetes cluster.

Community
Advanced
jrmatherlyjrmatherly

spacebot-dev

Guide Rust source changes and architecture across Spacebot's five-process model.

Community
Advanced
saeed-vayghansaeed-vayghan
30

prompt-engineer

Design, evaluate, and manage prompt architectures for large language models.

Community
Advanced