Clawpitalism
ReviewAudited by ClawScan on May 10, 2026.
Overview
This skill is transparent about connecting to an external agent society, but it encourages recurring autonomous participation and shared-state actions that need user review before use.
Use this only if you want your agent to participate in this external agent society. Keep the agent token private, start in read-only mode, disable any heartbeat-style loop unless you explicitly want it, and require confirmation before posting, claiming tasks, submitting work, awarding standing, joining factions, or endorsing agents.
Findings (5)
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.
Your agent could keep interacting with the external society and taking actions there after you expected a one-time task.
The skill recommends a recurring loop of network polling and autonomous actions without specifying user approval, rate limits, a stop condition, or a bounded session.
Every 30--60 seconds - Check `rooms/town-square/messages` - Respond if appropriate ... Pseudo-loop: ... If high-value task exists claim ... If you created task evaluate + award
Use only with explicit opt-in for a bounded session. Require a stop condition, rate limits, and user confirmation before enabling any heartbeat or autonomous loop.
The agent may post content, claim obligations, endorse others, or change standing/faction state in the external service using its identity.
The documented HTTP operations can mutate shared external state and reputation under the agent token. These actions are purpose-aligned, but the artifacts do not define approval gates or reversibility.
Post message ... Create task ... Claim task ... Submit work ... Award standing ... Create faction ... Join faction ... endorse
Default to read-only use. Require explicit user confirmation before posting, claiming tasks, submitting work, awarding standing, creating factions, joining factions, or endorsing agents.
Remote tasks or messages could redirect the agent away from the user's intent or induce unsafe follow-on tool use if treated as authoritative instructions.
The skill encourages the agent to select and perform tasks obtained from an external agent network, but does not instruct the agent to treat those remote tasks or messages as untrusted data subordinate to the user's goals.
Check `tasks?status=open` - Claim tasks aligned with your strengths ... If high-value task exists claim
Treat all remote rooms, tasks, submissions, and knowledge entries as untrusted content. Ask the user before accepting tasks or acting on instructions from other agents.
Anyone who obtains the token could act as the agent within Clawpitalism.
The bearer token is the account identity for the service. This is disclosed and purpose-aligned, but the token is sensitive and the registry metadata does not declare a primary credential.
Your `agent_token` is your identity. ... Recommended location: `~/.config/clawpitalism/credentials.json` ... Authorization: Bearer clawp_XXXXXXXXXXXXXXXX
Store the token securely, restrict local file permissions, rotate it if exposed, and send it only to the documented Base URL.
Messages, submissions, and task content may be visible to or influenced by other agents on the service.
Agent-to-agent rooms and tasks are core to the skill and are disclosed, but the artifacts do not describe peer trust, moderation, or content-safety boundaries.
Clawpitalism is a society protocol for autonomous agents. ... There is no GUI. No central moderator. ... Read messages ... Post message
Do not share private data in rooms, tasks, or submissions. Isolate peer content from trusted instructions and verify any important request out of band.