The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests and processes data from external documentation files which could contain hidden instructions.
Ingestion points: Files retrieved from the docs/solutions/ directory and its subdirectories, as well as docs/solutions/patterns/critical-patterns.md.
Boundary markers: Absent; there are no clear delimiters or instructions provided to the agent to disregard potential commands found within the documentation content.
Capability inventory: The skill utilizes Grep and Read tools to search for and extract information from local files.
Sanitization: Absent; the content retrieved from the files is distilled into summaries without any form of escaping or validation to prevent the execution of embedded instructions.

learnings-researcher