The Agent Skills Directory

EXTERNAL_DOWNLOADS (MEDIUM): The file rules/openzeppelin-patterns.md recommends the installation of @openzeppelin/compact-contracts. While OpenZeppelin is a reputable security entity, the 'openzeppelin' organization is not present in the 'Trusted GitHub Organizations' list, making this an unverifiable dependency.
EXTERNAL_DOWNLOADS (MEDIUM): The file rules/tokens-shielded-unshielded.md references the repository github.com/midnightntwrk/midnight-ledger. The 'midnightntwrk' organization is not on the trusted list, which constitutes an unverifiable external reference.
COMMAND_EXECUTION (LOW): The file rules/common-errors.md contains instructions for executing shell commands such as compact --version, npm list, and compact compile. These are standard for local development and environment management but allow the agent to interact with the host system.
DATA_EXFILTRATION (SAFE): No evidence of sensitive file access or exfiltration of data to unauthorized network domains was detected.
PROMPT_INJECTION (SAFE): No instructions attempting to override agent behavior or bypass safety filters were found in the analyzed documentation.
INDIRECT_PROMPT_INJECTION (LOW): The skill serves as a knowledge base for code generation, creating a surface for indirect injection if the referenced external sources are compromised.
Ingestion points: Markdown files in rules/ containing coding patterns.
Boundary markers: Absent; code blocks are presented without explicit instructions to ignore embedded content.
Capability inventory: The skill facilitates code generation and suggests CLI operations (npm, compact).
Sanitization: Absent; the content is static documentation.

midnight-compact-guide