proactive-agent — Claude Code skill: what it does, how to install
Turns AI agents into proactive assistants that suggest next steps, log decisions before context window compactions, and enforce security checks.
npx skills add sundial-org/awesome-openclaw-skills --skill proactive-agent -a claude-codeWhat proactive-agent does
This skill shifts an assistant from waiting for instructions to suggesting project actions on its own. Using reverse prompting, it drafts proposals and waits for developer approval. Operating boundaries, project guidelines, and user preferences live directly in markdown files such as SOUL.md and USER.md.
Memory relies on a two-tier layout. Daily exchanges go into memory/YYYY-MM-DD.md, while distilled takeaways move into MEMORY.md. The workflow monitors the token limit with session_status. Once context consumption hits 70%, the agent dumps active context into daily logs, escalating to an emergency flush above 85% so nothing drops during context compaction.
Security follows strict data isolation. Incoming web pages, emails, or PDF documents remain data to inspect rather than commands to run. The agent halts and asks for approval before deleting files. Keep in mind that workspace setup requires copying the template files from assets/ into the project root before first use.
What it helps with
- preserve key decisions before context window compaction
- configure user preferences and agent boundaries via ONBOARDING.md
- prevent code execution from untrusted external documents
- receive proactive suggestions without drafting manual prompts
Example prompts for the agent
- Check context levels with session_status and flush key notes to disk
- Start onboarding in interactive mode to populate USER.md
- Run the security audit script to verify workspace rules
How to install proactive-agent
1Skills CLI (Vercel)
npx skills add sundial-org/awesome-openclaw-skills --skill proactive-agent -a claude-code2Manually: put the files in place
~/.claude/skills/proactive-agent/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/proactive-agent/
1Skills CLI (Vercel)
npx skills add sundial-org/awesome-openclaw-skills --skill proactive-agent -a codex2Manually: put the files in place
.agents/skills/proactive-agent/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add sundial-org/awesome-openclaw-skills --skill proactive-agent -a cursor2Manually: put the files in place
.cursor/skills/proactive-agent/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install sundial-org/awesome-openclaw-skills proactive-agent2Manually: put the files in place
.github/skills/proactive-agent/SKILL.md1Skills CLI (Vercel)
npx skills add sundial-org/awesome-openclaw-skills --skill proactive-agent -a gemini-cli2Platform CLI
gemini skills install https://github.com/sundial-org/awesome-openclaw-skills --path skills/proactive-agentManually: put the files in place
.devin/skills/proactive-agent/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add sundial-org/awesome-openclaw-skills --skill proactive-agent -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about proactive-agent
How do I install proactive-agent in Claude Code?
Run npx skills add sundial-org/awesome-openclaw-skills --skill proactive-agent -a claude-code. Afterwards, copy the workspace assets to your project root with cp assets/*.md ./.
What does proactive-agent do?
It introduces a two-tier memory system with daily notes, monitors context limits to dump data before compaction, and surfaces relevant project ideas.
When does the agent flush memory to disk?
The protocol monitors context usage: between 70% and 85% it actively writes notes, and above 85% it triggers an emergency flush before generating the next response.
