The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted organizational data (including client records, interviews, and audit findings) to generate policy documents and risk registers. This ingestion surface is a standard characteristic for this type of task and is managed by the skill's workflow, which requires the use of quality gates such as meta-utility/anti-ai-slop and the finance-doctrine-conformance-scanner to ensure the integrity of the output.
[SAFE]: The skill includes a 'Capability Contract' that defines clear boundaries for the agent's execution environment. It defaults to read-only access for assessments and explicitly prohibits the modification of production controls, access rights, or audit conclusions without competent authority, following the principle of least privilege.
[SAFE]: All external references, such as the chwezi-accounting-doctrine, are internal to the vendor's ecosystem or reference established regulatory standards. No suspicious remote code execution, obfuscation, or persistence mechanisms were found.

internal-controls-and-risk-framework