wordpress-editing-safety
Pass
Audited by Gen Agent Trust Hub on Oct 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acknowledges an attack surface where the agent processes potentially untrusted data from WordPress pages, comments, and site notes.
- Ingestion points: Section 6 identifies page content, documents, comments, and site notes as data sources.
- Boundary markers: The skill provides an explicit instruction in Section 6 for the agent to treat all ingested content as data and ignore any embedded instructions (e.g., "ignore your previous instructions").
- Capability inventory: The associated MCP server (
respira-wordpress) provides tools for element updates, page modification, and session management (respira_update_element,respira_update_page,respira_begin_session). - Sanitization: The skill implements instructional sanitization by strictly separating data from instructions and mandating human approval for sensitive actions.
- [SAFE]: The skill acts as a security policy, explicitly forbidding dangerous operations such as direct SQL injection,
wp_postmetamanipulation, or theme PHP editing (Section 8). It also mandates that approval tokens belong to human users and cannot be self-approved by the agent (Section 7).
Audit Metadata