The Agent Skills Directory

[SAFE]: The skill is designed for standards compliance assessment and does not contain any executable code or scripts that perform dangerous operations. All instructions are focused on analysis, citation of standards, and reporting findings.
[PROMPT_INJECTION]: No attempts to override agent safety guidelines, bypass filters, or extract system prompts were detected. The instructions reinforce existing safety boundaries, such as forbidding direct code modification.
[DATA_EXFILTRATION]: No network operations to unknown domains or commands designed to harvest sensitive local files (e.g., .ssh, .aws credentials) were found. The skill logs results locally to the .agents/ directory as part of standard project tracking.
[REMOTE_CODE_EXECUTION]: The skill does not download or execute remote scripts. While it references various third-party security and accessibility testing tools in its documentation (e.g., axe-core, pa11y, npm audit), these are presented as recommendations for the user/agent to use in a standard auditing capacity.
[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to ingest and analyze external codebase data. While this presents a surface for indirect prompt injection, the skill mitigates risk by operating as an assessment-only agent and delegating all implementation tasks to other agents. The risk is assessed as low and inherent to the intended use-case of a code auditor.
[OBFUSCATION]: All files consist of plain-text markdown and YAML. No encoded content, hidden characters, or homoglyph-based obfuscation techniques were detected.

canon