The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection due to its data ingestion design.\n
Ingestion points: The skill reads from .claude/context/idea.md and .claude/context/build.md, and references multiple JSON datasets such as web3-ideas-combined.json and yc-requests-for-startups.json located in the ../ext/solana-new/ directory.\n
Boundary markers: There are no explicit instructions or delimiters provided to the agent to help distinguish between its own system instructions and potentially untrusted content from the ingested files.\n
Capability inventory: The skill has the capability to modify the local filesystem by writing or updating the .claude/context/idea.md file.\n
Sanitization: The skill instructions do not specify any validation, filtering, or escaping of the content retrieved from external files before it is processed.

idea-sprint