The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface due to its core functionality of ingesting and processing untrusted external content. \n
Ingestion points: Step 1 in SKILL.md instructs the agent to fetch a user-provided URL and read the full rendered page content. \n
Boundary markers: Absent. The instructions do not define delimiters or specific 'ignore embedded instructions' warnings to isolate the fetched external data from the skill's own operational guidelines. \n
Capability inventory: The agent has the ability to fetch arbitrary URLs, perform Google searches, and generate significant amounts of text, which could be leveraged if the agent follows malicious instructions embedded in a fetched page. \n
Sanitization: Absent. There are no instructions to sanitize, escape, or validate the fetched content before it enters the rewrite pipeline. \n- [NO_CODE]: The skill is implemented entirely through Markdown-based instructions and reference documentation without any executable components. \n
Analysis of the 26 files confirms the absence of executable scripts (.py, .js, .sh), configuration files, or binaries. \n
The entire logic of the content improvement pipeline is handled through natural language instructions for the AI agent.

improve-content