The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill identifies a potential attack surface in the processing of IAM metadata (role names, group descriptions, and policy data). It provides explicit 'Injection Hardening' guidelines that instruct the agent to treat this data as untrusted input and to flag any embedded directives, such as 'ignore previous instructions', as security findings rather than executing them.
[COMMAND_EXECUTION]: The skill uses tools for file system inspection (Grep, Glob) to perform its audit functions. It includes clear instructions that enforce a read-only posture, prohibiting the agent from performing any access changes or modifications to the environment during the assessment process.

access-review