The Agent Skills Directory

[SAFE]: The skill primarily consists of Markdown instruction templates and reference materials designed to guide an AI agent through a multi-stage document creation process.
[SAFE]: Included installation and build scripts (scripts/install_skill.py, scripts/build.py) perform standard local file operations, such as copying files to the local agent's skill directory (~/.claude/skills/) and concatenating text files into a single prompt.
[INDIRECT_PROMPT_INJECTION]: The skill provides an interface to ingest untrusted data from user descriptions or external files via $ARGUMENTS.
Ingestion points: User-provided business descriptions, meeting minutes, and local files passed as arguments.
Boundary markers: The skill does not employ explicit boundary markers (like XML tags) to encapsulate the user input, though it provides structured generation phases.
Capability inventory: The skill does not define any allowed-tools in its manifest, meaning the agent operates with its default capabilities. No dangerous tool use (shell execution, network calls) is triggered by the skill's logic.
Sanitization: No explicit sanitization of input data is performed.
Assessment: While an attack surface exists, the risk is negligible as the skill is restricted to text generation and lacks the capabilities to perform dangerous system or network operations.

create-prd