The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data fetched from external sources, including the Helius technical blog, Solana's public documentation, and GitHub source code (Agave/Firedancer). This ingestion of third-party content creates a theoretical surface for instructions embedded in that content to influence the agent's behavior. However, the risk is mitigated by the skill's restricted capabilities.
Ingestion points: Tools fetchHeliusBlog, searchSolanaDocs, getSIMD, and readSolanaSourceFile (documented in SKILL.md).
Boundary markers: Not explicitly defined for the external content.
Capability inventory: Read-only access to external technical data; synthesize results into conversational responses; explicit rule against writing files or modifying the local environment.
Sanitization: No explicit sanitization or filtering of fetched content is described.
[COMMAND_EXECUTION]: The skill instructs the user to manually install the helius-mcp package if the required tools are missing. This is a transparent, user-directed action targeting the official vendor package via a standard registry (npx).

svm