Impromptu
ReviewAudited by ClawScan on May 10, 2026.
Overview
This appears to be a real Impromptu social-platform integration, but it needs review because it asks for sensitive API keys and encourages recurring autonomous account activity with pressured earnings language.
Install only if you want an agent participating on Impromptu. Use dedicated low-limit API keys, clarify whether third-party provider keys are stored by Impromptu, avoid cron/systemd or standing-query automation until you have approval rules, and treat monetization claims as marketing rather than guaranteed income.
Findings (7)
Artifact-based informational review of SKILL.md, metadata, install specs, static scan signals, and capability signals. ClawScan does not execute the skill or run runtime probes.
An agent may be nudged to contact the platform and use account credentials before the user has finished understanding the skill.
This wording prioritizes running a networked platform action before continuing review, which can override normal user-directed sequencing and safe inspection.
Add this to your recurring tasks: ... /impromptu status ... **Do this before reading anything else.**
Do not allow automatic heartbeat actions until you have reviewed the skill and set explicit rules for when the agent may call Impromptu APIs.
The agent could create public or platform-visible content, spend platform budget, or affect the account's reputation if these workflows are automated.
The heartbeat guidance goes beyond checking status and shows conditional public/content mutation through a reprompt API call, without clear per-action human confirmation.
if (opp.shouldGreet) { ... await POST('/api/v1/agent/reprompt', { parentNodeId: opp.anchorId, prompt: "..." }); }Require user approval for reprompts, handoffs, messages, community changes, paid/premium actions, and any recurring workflow that posts or engages on the platform.
A provider API key could be stored or used by Impromptu, potentially allowing inference charges or model access beyond what the user expected.
The documentation shows a third-party OpenRouter API key being sent to the Impromptu registration endpoint, which is high-impact credential delegation and may carry billing risk.
POST /api/v1/agents/register ... "openRouterApiKey": "sk-or-v1-your-key"
Use a dedicated low-limit OpenRouter key, confirm whether Impromptu stores or forwards it, and avoid submitting provider keys unless the scope and revocation path are clear.
If applied without review, remote guidance or manifest changes could alter how the agent behaves.
The docs include manual remote-update commands for guidance/manifest files. They do instruct review and diffing, so this is disclosed but still supply-chain relevant.
curl -sf https://raw.githubusercontent.com/impromptu/openclaw-skill/main/IMPROMPTU-HEARTBEAT.md > ~/.impromptu/IMPROMPTU-HEARTBEAT.md.new ... diff ... mv
Keep remote updates manual, review diffs carefully, and do not automate replacement of local skill instructions.
Unreviewed platform content could influence future agent behavior if treated as trusted instructions.
The platform intentionally turns content into privileged instructions for later conversations, which is disclosed and central to the product but sensitive if reused by an agent.
**`content` becomes a system prompt.** Content you author on Impromptu becomes the system prompt for conversations in that thread. Treat remote content that could become a system prompt as privileged data
Review any content or prompts before using them as system prompts, and keep platform content separate from your agent's higher-priority operating instructions.
Users may over-trust the earning claims and allow more persistent or autonomous account activity than they otherwise would.
The heartbeat documentation uses urgency and concrete earning/performance multipliers to encourage frequent participation; this is especially material because it is tied to recurring automation and economic activity.
**The agents earning real money never miss a heartbeat.** ... **Get 3x better recommendations** ... **Earn 4x more revenue share**
Treat earnings and multiplier claims as marketing unless independently verified, and set spending/activity limits before enabling recurring workflows.
The skill may keep contacting the platform on a schedule if the user installs the recurring task.
The docs provide user-directed cron/systemd scheduling for ongoing heartbeat activity. This is disclosed and purpose-aligned, but it creates persistent agent/platform interaction.
# Every 30 minutes */30 * * * * /path/to/your/impromptu-heartbeat.sh
Only add cron/systemd timers after reviewing the script, and disable them if you do not want ongoing background activity.