Pilot Swarm Config

Distributed configuration management for agent swarms with versioned updates. Use this skill when: 1. Multiple agents need to share configuration settings 2....

This skill appears to do what it says (distributed, versioned config via pilotctl) but the SKILL.md assumes variables and dependencies that are not declared in the registry metadata. Before installing: (1) Verify pilotctl and jq are trusted binaries on the host and confirm how pilotctl authenticates to the registry (credentials may be stored outside the skill). (2) Confirm what values will be used for SWARM_NAME, AGENT_ID, and REGISTRY_HOST and whether those identifiers expose sensitive mapping between agents and identities. (3) Test in an isolated environment to observe what the skill publishes to the registry and whether the registry endpoint is trusted. (4) Ask the publisher to update metadata to declare required env vars (SWARM_NAME, AGENT_ID, REGISTRY_HOST), the pilot-protocol dependency, and any authentication needs. If you cannot confirm those details, avoid granting this skill autonomous invocation on production agents.

Type: OpenClaw Skill Name: pilot-swarm-config Version: 1.0.0 The skill contains multiple shell injection vulnerabilities in SKILL.md due to the use of unquoted and unsanitized variable interpolation within Bash commands. Specifically, variables like $SWARM_NAME, $CONFIG_DATA, and $CONFIG_CHANNEL are passed directly into pilotctl arguments and jq filters (e.g., in the 'Apply configuration locally' and 'Workflow Example' sections), which could allow for command execution if the agent processes untrusted input. While the functionality aligns with the stated purpose of configuration management, these implementation flaws pose a significant security risk.