The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external web sources which could potentially contain malicious instructions.
Ingestion points: External web content retrieved via the /sc:research tool and local project configuration files such as package.json, Cargo.toml, and go.mod in SKILL.md.
Boundary markers: The instructions lack explicit delimiters or boundary markers to isolate external research data from the agent's primary instructions, increasing the risk that embedded commands in web content might be obeyed.
Capability inventory: The skill possesses capabilities to read local files, execute the /sc:research slash command, and write new documentation files to the claudedocs/ directory.
Sanitization: There are no instructions for sanitizing or validating the content returned by the research tool before it is synthesized into a report.

research-guide