The Agent Skills Directory

[DATA_EXPOSURE_EXFILTRATION]: The skill uses web_search and image_search to verify facts and find visual aids. These are legitimate tools for its stated purpose of rewriting content with 'real facts'. No sensitive data access or unauthorized exfiltration patterns were detected.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied social media posts (untrusted data) as its primary input. While this creates an attack surface where a user might attempt to inject instructions into the search query, the skill is instructed to extract the 'core meaning' and search for objective facts, which naturally filters most prompt injection attempts.
Ingestion points: SKILL.md (user-provided tweet/post content)
Boundary markers: Absent
Capability inventory: web_search, image_search
Sanitization: None specified, but logic focuses on extracting factual entities for searching.
[COMMAND_EXECUTION]: No shell commands, subprocess calls, or privilege escalation attempts were found. The skill operates entirely within the context of text manipulation and standard search tool calls.

wjs-polishing-x-engagement