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/primeline-ai/evolving-lite/evolution-guidenpx skills add primeline-ai/evolving-lite --skill evolution-guidegit clone --depth 1 https://github.com/primeline-ai/evolving-liteWhat 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 | $0.00057 | $0.00869 |
| Opus 5 | $0.00028 | $0.00434 |
| Sonnet 5 | $0.00011 | $0.00174 |
| Haiku 4.5 | $0.00006 | $0.00087 |
Grade B, and why
evolution-guide scanned grade B with 1 finding 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.
Reads agent configuration directoriesmediumAgent snooping
.claude/, .codex/, .gemini/ hold keys, settings and other credentials a mod has no legitimate need for.
Remove the plugin path from `pluginDirectories` in `~/.claude/settings.json`. Your memory data stays in the plugin directory - delete it if you want a clean removal, or keep it if you might reinstall later. How it starts
The opening of the file, as written. The whole thing — 74 lines — stays where its author put it; the contents beside it link to each section on GitHub.
Evolution Guide
When the user asks about Evolving Lite, explain based on their question:
What is Evolving Lite?
A Claude Code plugin that learns from your corrections, manages your context budget, and improves every session - automatically. You install it once and it works in the background while you code normally.
The 4 Feedback Loops
LEARN Loop (active from Tier 2, session 3+)
You correct Claude -> correction-detector recognizes it -> creates an Experience file -> next time a similar problem comes up, thinking-recall injects that experience mid-conversation. Claude doesn't make the same mistake twice.
HEAL Loop (active from Tier 1)
health-sentinel checks at every session start whether all hooks ran correctly last session. If a hook silently failed (no sentinel output), you get a warning immediately. The system also has integrity-checker/fixer agents for deeper repairs.
EVOLVE Loop (active from Tier 1)
usage-tracker counts what tools and commands you use. auto-archival removes stale experiences (>90 days unused). The system gets leaner and more focused over time without you doing anything.
CONTEXT Loop (active from Tier 1)
context-warning monitors your context budget. At 70% it warns you. At 93% it suggests creating a handoff. precompact-extract saves important knowledge before compaction wipes it. You never lose context silently.
Tiered Activation
All hooks are installed from day one but activate progressively:
| Tier | Active from | What activates | Purpose |
|---|---|---|---|
| 1 | Session 1 | context-warning, security-tier-check, health-sentinel, usage-tracker | Safety & monitoring |
| 2 | Session 3 | correction-detector, delegation-enforcer, session-summary | Learning & delegation |
| 3 | Session 10 | thinking-recall, auto-archival, precompact-extract | Deep memory & maintenance |
Available Commands
All commands are optional. The system works fully automatically without using any of them.
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 · 74 lines · 57 tokens per session scan B 650b545e8a31
evolution-guide is a skill published in the GitHub repository primeline-ai/evolving-lite (48 stars, last pushed 16d ago), licensed MIT. It adds 57 tokens to every session and 869 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it B with 1 finding (reads agent configuration directories). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other skills, from other repositories
subagent-strategy
Delegates research and parallel work to sub-agents.
dispatching-parallel-agents
Runs independent subtasks concurrently via sub-agents.
spawn-agent
Spawn worker agents (Gemini CLI or Codex CLI) to keep main context clean. Use for implementation, codebase research, context gathering, or any scoped work that would pollute the orchestrator's context.
skill-dispatcher
Universal MCP skill dispatcher for AI coding agents. Loads Claude Code, OpenCode, Cursor, Gemini-style, command, plain markdown, and GitHub-hosted skills on demand. Includes agent routing, active skill lifecycle tracking, external repo import, nested skill discovery, origin filtering, and hardened parsing for secure…
extension-creator
Create, scaffold, customize, validate, and locally test Qwen Code extensions. Use when the user wants a new Qwen Code extension, needs help choosing an extension template, wants to add QWEN.md context, commands, skills, agents, MCP servers, settings, hooks, channels, or LSP servers, or asks how to link and test an…
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.