The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted external data including JIRA exports, Linear tickets, PRDs, Git logs, and content from product URLs.
Ingestion points: Instructions specifically direct the agent to read user-provided files and perform web searches on product URLs in SKILL.md.
Boundary markers: The instructions lack explicit delimiters or warnings to the agent to ignore instructions that might be embedded within the processed tickets or documents.
Capability inventory: The skill has the capability to read local files and access the internet via web search.
Sanitization: There is no evidence of sanitization or validation of the content retrieved from external sources before it is processed by the agent.
[PROMPT_INJECTION]: The skill uses the $ARGUMENTS variable directly in the context section: 'You are writing release notes for $ARGUMENTS'. This allows user-supplied text to be interpolated directly into the system instructions, which could be used to attempt to influence the agent's behavior if the input is crafted maliciously.

release-notes