The Agent Skills Directory

[NO_CODE]: The skill consists entirely of markdown instructions for behavior and persona adoption. It does not contain any scripts, binaries, or command-line executions.
[INDIRECT_PROMPT_INJECTION]: The skill defines a process for ingesting information from workspace files to parameterize its responses. Ingestion points: The instructions explicitly require reading the workspace's README.md, account/operator.md, and users.md to understand business context. Boundary markers: There are no specific delimiters or instructions provided to the agent to distinguish between the business data and potential embedded instructions within those files. Capability inventory: The skill itself possesses no capabilities for file writing, network communication, or system command execution across its documentation. Sanitization: No sanitization or validation logic is defined for the ingested data. Impact: Given the total absence of executable tools or system-level access within the skill, the risk of an indirect prompt injection attack causing harm is negligible.

cpo