The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The file 'references/submission-tracker-template.csv' is identified as malicious by file reputation scanners (FileRepMalware).
[EXTERNAL_DOWNLOADS]: The directory reference list in 'references/directory-list.md' contains a link to 'https://fitprofessionals.net', which is currently blacklisted by automated security scanners.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it downloads and processes untrusted HTML data from external websites to verify backlinks.
Ingestion points: Content is ingested from third-party directories using the 'curl' tool as described in 'SKILL.md'.
Boundary markers: The instructions explicitly warn the agent to treat directory pages as untrusted data and ignore embedded instructions, though this does not fully mitigate the risk.
Capability inventory: The agent has access to 'curl' and 'dig' for network operations and is instructed to use the 'mcp-publisher' CLI tool.
Sanitization: There is no automated sanitization of the fetched HTML content; the skill relies on manual human verification before proceeding with tasks.
[COMMAND_EXECUTION]: The skill utilizes shell-based tools like 'curl' and 'dig' to interact with external network resources for verification purposes, which can be exploited if the target hosts are malicious.

directory-submissions