The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves inspecting and editing existing README files and package metadata. This creates an attack surface where malicious instructions embedded in a project's README could potentially influence the agent's behavior. * Ingestion points: File inspection during 'Agent workflow' step 1 in SKILL.md. * Capability inventory: Modifies local README files and constructs external URLs. * Boundary markers: None present in instructions. * Sanitization: No explicit sanitization of README content is mentioned.
[DATA_EXFILTRATION]: The skill utilizes endpoints (such as '/badge/dynamic/json?url=...' and '/https/{hostname}/{path}') that allow fetching data from arbitrary external URLs. This functionality could be leveraged to exfiltrate information by causing the agent to include sensitive data in a request to the badge service's proxy.
[COMMAND_EXECUTION]: The documentation provides installation commands using npx (e.g., 'npx skills add jal-co/shieldcn'). While these are standard for skill management, they rely on the execution of remote packages from the specified author.

shieldcn-badges