The Agent Skills Directory

[SAFE]: The skill operates as a high-level instructional framework for document generation. It does not include executable code, scripts, or commands that interact with the system at a low level. All references to external modules, such as the chwezi-accounting-doctrine, are internal to the vendor's ecosystem and follow legitimate resource patterns.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and audit untrusted client data (e.g., existing HR policies, contracts, and staff feedback), which creates a theoretical attack surface for indirect prompt injection. However, this is necessary for the skill's primary function and is mitigated by structured workflows and quality control gates.
Ingestion points: The skill reads external HR policies and organizational records to perform gap audits and drafting as defined in the SKILL.md Inputs section.
Boundary markers: The references/document-blueprint.md implements a 'parameterization rule' that structurally separates client-supplied data from the static policy framework, and uses a Statutory Schedule to isolate volatile regulatory data.
Capability inventory: The skill's capabilities are restricted to reading text files and generating document artifacts (DOCX/PDF) through standard agent tools. There are no network operations, subprocess executions, or file-write capabilities targeting sensitive system areas.
Sanitization: The workflow includes specific steps to run output through quality gates, such as meta-utility/anti-ai-slop and meta-strategy/meta-critical-thinking-business-logic, to ensure the final output is logical and free from unintended instructions.

hr-policy-manual