The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructions include hardcoded, absolute Windows filesystem paths: C:/Users/cmano/claude-seo/.... This exposes a specific local username (cmano) and the host's directory structure to the agent. Attempting to write to or read from specific user profile directories is a form of sensitive data exposure and poses a risk to the host environment's privacy and security.
[PERSISTENCE]: The skill contains explicit instructions to modify third-party skill files (specifically seo-audit and seo from the AgriciDaniel/claude-seo repository). It directs the agent to inject logic into these files to trigger its own execution (spawn condicional and orchestration steps). Modifying other instruction files to ensure the skill remains active or is automatically invoked constitutes an application-level persistence mechanism.
[INDIRECT_PROMPT_INJECTION]: The core functionality of the skill involves using WebFetch to ingest untrusted data from external websites (Homepage, Privacy Policy, and Contact pages) for analysis. This creates a significant surface for indirect prompt injection attacks.
Ingestion points: External website content (HTML, headers, policies) fetched during 'Paso 1'.
Boundary markers: The skill does not define specific delimiters or instructions for the agent to ignore embedded commands within the fetched data.
Capability inventory: The skill has the capability to write files to the local disk and perform further network operations.
Sanitization: There is no evidence of sanitization or validation of the external content before it is processed by the agent.

ley-datos-chile