The Agent Skills Directory

[Prompt Injection] (SAFE): No instructions attempting to override agent behavior or bypass safety filters were found.
[Data Exposure & Exfiltration] (SAFE): No hardcoded credentials, sensitive file paths, or network requests were identified.
[Obfuscation] (SAFE): No hidden characters, Base64 encoding, or homoglyphs were detected.
[Unverifiable Dependencies & Remote Code Execution] (SAFE): No remote scripts are downloaded or executed, and no external packages are required.
[Privilege Escalation] (SAFE): No use of sudo or commands targeting system-level permissions.
[Persistence Mechanisms] (SAFE): No attempts to modify startup scripts or system services were found.
[Metadata Poisoning] (SAFE): Metadata fields are accurate and reflect the documentation purpose of the skill.
[Indirect Prompt Injection] (SAFE): The skill does not ingest untrusted external data that could influence the agent context.
[Time-Delayed / Conditional Attacks] (SAFE): No conditional logic based on time or environment triggers.
[Dynamic Execution] (SAFE): No dynamic code generation or unsafe deserialization patterns.

grepai-chunking