The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it is designed to ingest and analyze untrusted external data (legal contracts).
Ingestion points: Processes external contract documents and project files to perform redlining and risk scoring.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potential malicious commands embedded within the contracts being reviewed.
Capability inventory: The agent has access to powerful tools including Bash, Write, and WebFetch, which could be targeted by an injection attack within a contract.
Sanitization: There are no instructions for sanitizing or escaping content from the ingested documents before processing.
[COMMAND_EXECUTION]: The skill requests access to the Bash tool in its frontmatter. While the current instructions use it for standard file operations (Grep, Glob), the presence of a shell environment allows for broader command execution if the agent is influenced by malicious input.
[EXTERNAL_DOWNLOADS]: The skill uses WebFetch and WebSearch to retrieve regulatory guidance. This is a legitimate use case for a legal review tool but involves connecting to external domains.

brief-review