The Agent Skills Directory

[PROMPT_INJECTION]: No patterns attempting to bypass safety filters, override agent instructions, or extract system prompts were found.
[DATA_EXPOSURE_AND_EXFILTRATION]: No sensitive file access or credential exposure detected. External links point exclusively to official Microsoft documentation (learn.microsoft.com), which is a trusted source.
[OBFUSCATION]: The content is clear and uses no encoding (Base64, hex), zero-width characters, or homoglyphs to hide instructions.
[UNVERIFIABLE_DEPENDENCIES_AND_REMOTE_CODE_EXECUTION]: No remote scripts are downloaded or executed. The skill references the requirement for a locally installed .NET SDK, which is consistent with its stated purpose.
[PRIVILEGE_ESCALATION]: No commands for escalating privileges (sudo, chmod) or modifying system-level configurations are present.
[PERSISTENCE_MECHANISMS]: No logic exists to modify shell profiles, cron jobs, or startup services for maintaining unauthorized access.
[METADATA_POISONING]: Metadata fields are descriptive and align with the technical content provided in the skill files.
[INDIRECT_PROMPT_INJECTION]: The skill analyzes local project files (e.g., .csproj, views, controllers) to provide context-aware advice. While this represents a standard ingestion point, the skill lacks the capabilities (such as network sending or subprocess execution) to be leveraged for malicious injection.
Ingestion points: Local project structure and source files (SKILL.md).
Boundary markers: None explicitly defined, relies on agent defaults.
Capability inventory: Purely instructional; no file-writing or network capabilities identified.
Sanitization: Not applicable as it is an advisory skill.
[DYNAMIC_EXECUTION]: No use of eval, exec, or runtime compilation of generated code detected.
[DYNAMIC_CONTEXT_INJECTION]: The skill does not use the load-time shell command execution syntax.

neo-dotnet-mvc