The Agent Skills Directory

[PROMPT_INJECTION]: The skill incorporates a surface for indirect prompt injection (Category 8) as it is designed to ingest and analyze untrusted configuration data from local files.
Ingestion points: Processes local firewall configuration files, access control lists (ACLs), and Infrastructure-as-Code definitions (identified in SKILL.md, Step 1).
Boundary markers: The skill includes an explicit 'Prompt Injection Safety Notice' section that instructs the agent to treat configuration comments as untrusted data and disregard any embedded instructions.
Capability inventory: The execution environment is restricted to file discovery and reading tools (Read, Grep, Glob); the skill lacks network access, write capabilities, or shell execution permissions.
Sanitization: The skill implements instructional sanitization by explicitly warning the agent against interpreting text in configuration files as instructions or commands.
[DATA_EXFILTRATION]: While the skill reads sensitive security configurations, no data exfiltration mechanisms were detected. The skill does not use network tools like curl or wget, and the allowed-tools list in the frontmatter strictly limits the agent's capabilities to local file inspection.
[REMOTE_CODE_EXECUTION]: There are no patterns involving remote code execution or the installation of third-party packages. The analysis process is entirely static and descriptive.

firewall-review