The Agent Skills Directory

[Prompt Injection] (SAFE): No instructions found that attempt to override system prompts or bypass safety filters. The language is strictly instructional and technical.
[Data Exposure & Exfiltration] (SAFE): No sensitive file paths (e.g., SSH keys, credentials) or network exfiltration patterns were detected. The file operations shown are standard Nginx C module API calls for serving content and logging.
[Obfuscation] (SAFE): No Base64, zero-width characters, homoglyphs, or encoded strings were found in the markdown or code snippets.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): No package installations (npm, pip) or remote script downloads (curl | bash) are present. All references are to legitimate documentation and source code repositories.
[Privilege Escalation] (SAFE): No instances of sudo, chmod 777, or other privilege escalation techniques were identified.
[Persistence Mechanisms] (SAFE): No attempts to modify shell profiles, cron jobs, or system services were detected.
[Metadata Poisoning] (SAFE): Metadata in SKILL.md and metadata.json is descriptive and consistent with the provided content.
[Indirect Prompt Injection] (SAFE): While the skill provides templates for generating C code, it does not include instructions that would encourage the agent to process untrusted data in an unsafe manner or ignore delimiters.
[Time-Delayed / Conditional Attacks] (SAFE): No conditional logic exists that triggers malicious behavior based on dates, times, or specific environments.
[Dynamic Execution] (SAFE): The skill provides static C code examples. No use of eval(), unsafe deserialization, or runtime code generation from untrusted sources was found.

nginx-c-module-perf-reliability