The Agent Skills Directory

[SAFE]: The skill demonstrates high security awareness by incorporating defensive measures against common data processing vulnerabilities. It explicitly instructs the agent to redact sensitive cryptographic fields such as JWTs, private keys, and entropy seeds during inspection and logging.\n- [INDIRECT_PROMPT_INJECTION]: While the skill is designed to ingest and process potentially untrusted ACVP JSON artifacts, it implements robust mitigations. The provided Python inspection script (scripts/inspect-acvp-json.py) enforces strict limits on input size (64MB), nesting depth (64), and aggregate member counts (1 million), which prevents resource exhaustion attacks. The Kotlin template (assets/AcvpEnvelope.kt) utilizes strict JSON configurations and validation checks for hex bit-lengths and integer ranges to ensure structural integrity.\n
Ingestion points: scripts/inspect-acvp-json.py (reading file paths) and assets/AcvpEnvelope.kt (parsing JSON strings).\n
Boundary markers: Not applicable as it processes structured data with specific schema validation.\n
Capability inventory: The skill's operations are limited to data classification, structural reporting, and modeling; it does not include tools for network communication or arbitrary command execution based on artifact content.\n
Sanitization: Implements sensitive field redaction and resource constraints as defensive layers.

libacvp-json-kotlin