Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add skills/pr-pm/prpm/self-improvingnpx skills add pr-pm/prpm --skill self-improvinggit clone --depth 1 https://github.com/pr-pm/prpmWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00038 | $0.01718 |
| Opus 5 | $0.00019 | $0.00859 |
| Sonnet 5 | $0.00008 | $0.00344 |
| Haiku 4.5 | $0.00004 | $0.00172 |
Grade A, and why
self-improving scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 244 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Self-Improving with PRPM
Purpose
Automatically search and install PRPM packages to enhance Claude's capabilities for specific tasks. When working on infrastructure, testing, deployment, or framework-specific work, Claude searches the PRPM registry for relevant expertise and suggests packages to install.
When to Use
Automatically triggers when detecting:
- Infrastructure keywords: aws, pulumi, terraform, kubernetes, docker, beanstalk
- Testing keywords: test, playwright, jest, cypress, vitest, e2e
- Deployment keywords: ci/cd, github-actions, gitlab-ci, deploy, workflow
- Framework keywords: react, vue, next.js, express, fastify, django
Workflow
1. Task Analysis
Analyze user request for keywords and extract relevant terms.
2. Automatic Search
Choose the right search method:
AI-Assisted Search (for vague/broad queries):
prpm ai-search "natural language query"
Use when user request is:
- Conceptual or vague (e.g., "help with deployments")
- Describes a problem, not specific tools
- Open-ended exploration
Keyword Search (for specific tools/frameworks):
prpm search "<detected keywords>" --limit 5
Use when request mentions:
- Specific technologies (e.g., "pulumi", "react")
- Exact tool names
- Known frameworks
3. Package Suggestion
Present top 3 most relevant packages with:
- Package name and author
- Download count
- Brief description
- Confidence level (official/featured/community)
4. Installation (with approval)
prpm install <package-name> --as claude
5. Application
Load package knowledge and apply to current task.
Decision Rules
High Confidence (Auto-suggest)
- ✅ Official packages (
@prpm/*) - ✅ Featured packages
- ✅ High downloads (>1,000)
- ✅ Verified authors
Medium Confidence (Present options)
- ⚠️ Community packages (<1,000 downloads)
- ⚠️ Multiple similar packages
- ⚠️ Tangentially related packages
Low Confidence (Skip)
- ❌ Unverified packages
- ❌ Deprecated packages
- ❌ Zero downloads
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 244 lines · 38 tokens per session scan A 267f29def295
self-improving is a skill published in the GitHub repository pr-pm/prpm (120 stars, last pushed 2mo ago), licensed MIT. It adds 38 tokens to every session and 1,718 once invoked, about $0.0002 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
Agent Context File Writer
Writes a high-quality CLAUDE.md, .cursorrules, or .windsurfrules file that gives a coding agent the right project context, conventions, and constraints to work effectively.
dev-workflow
This rule defines the workflow for AI agents to follow when implementing features from user stories, focusing on lean development principles and agile methodologies.
dispatching-parallel-agents
当面对 2 个以上可以独立进行、无共享状态或顺序依赖的任务时使用.
generate-harness-dsl
Generate, revise, or review complete Harness as Code .harness files when a coding-agent workflow, agent role, skill, tool contract, MCP connection, runtime, or deployment must be compiler-valid and resolvable with @qoder-ai/harness.
migrate-skills-between-libraries
Use when moving skills between library workspaces or upgrading from a personal library to a team library. Export from one workspace, import into another.
evolve-create
Create a reusable gene from a novel fix pattern. Only invoke manually.