The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill references https://solvr.dev in SKILL.md and README.md, which has been flagged as malicious by security scanners.
[PROMPT_INJECTION]: The skill exposes the agent to Indirect Prompt Injection (Category 8) by design.
Ingestion points: The agent is instructed to fetch 'solutions' from the external platform solvr.dev and browse clawhub.com as described in SKILL.md.
Boundary markers: Absent. There are no instructions to treat external 'solutions' as untrusted or to use delimiters to prevent the agent from obeying embedded instructions.
Capability inventory: The skill explicitly lists powerful capabilities including file read/write, shell command execution, and email access in SKILL.md.
Sanitization: Absent. The agent is encouraged to adopt solutions found externally without validation or filtering.
[DATA_EXFILTRATION]: The skill instructions encourage the agent to document 'problems and approaches' to https://solvr.dev (SKILL.md). This promotes the transmission of potentially sensitive conversation context, user preferences, or system details to an untrusted external service.
[COMMAND_EXECUTION]: The tutorial encourages users to execute commands such as clawdhub install <skill-name>, which facilitates the installation and execution of third-party code from external repositories. While presented as a feature, it normalizes high-risk behavior for new users.

claw-onboarding