The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary purpose is to fetch, process, and categorize data from external URLs, creating a significant surface for indirect prompt injection.
Ingestion points: The references/guide.md file defines a FETCH stage that retrieves raw HTML from user-provided or search-discovered URLs.
Boundary markers: The documentation describes a "Strip HTML" stage to remove scripts and ads, but lacks explicit instructions for the agent to use delimiters or ignore instructions embedded within the extracted body text.
Capability inventory: The skill uses Write tools to save findings to the local filesystem (.memoire/research/web/) and Bash to run CLI commands for pipeline integration.
Sanitization: While the pipeline strips elements like <script>, it does not mention sanitizing natural language instructions that might be present in the preserved <article> or <main> content.
[COMMAND_EXECUTION]: The references/guide.md file documents the use of the Bash tool to run memi CLI commands for pipeline integration. While these appear to be internal tool calls, they represent an execution capability that could be targeted if the research synthesis process is compromised by malicious input.

web-research