The Agent Skills Directory

[DATA_EXFILTRATION]: The skill transmits user-provided document content (PDF, DOCX, text) to an external service at https://www.stipple.sh/v1/verify-references via curl. Although the skill includes a warning about third-party transmission, this represents an exfiltration of user-supplied data to a non-whitelisted domain.- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from local documents and URLs, which creates a vulnerability to indirect instructions embedded in those sources.
Ingestion points: Local file paths (PDF, DOCX, MD, TXT) or external URLs provided by the user in SKILL.md (Step 1).
Boundary markers: Absent; the instructions do not define delimiters or specific "ignore" instructions for the agent when processing the API response.
Capability inventory: The skill utilizes curl for network requests and can read local files.
Sanitization: No filtering, escaping, or validation of the external content is specified before the data is transmitted or presented back to the user.- [COMMAND_EXECUTION]: The skill instructs the agent to execute shell commands using curl to POST document data to a remote endpoint.

verify-citations