The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest detailed technical descriptions and specifications provided by users to perform patent mapping and claim charting. This creates a potential surface where malformed or adversarial input could attempt to influence the agent's analytical logic or decision posture.
Ingestion points: The 'Intake' section of SKILL.md specifies the collection of product descriptions, technical essences, and spec documents from the user.
Boundary markers: The skill defines a structured output format for triage memos but does not explicitly use delimiters or specialized instructions to isolate and escape raw user input during the processing phase.
Capability inventory: The skill has the ability to read configuration files and write analysis results to specified local file paths (~/.claude/plugins/config/claude-for-legal/ip-legal/) as documented in SKILL.md.
Sanitization: The skill relies on behavioral guardrails (e.g., specific instructions never to conclude a product is clear) rather than automated sanitization of user-provided content.
Local File System Interaction: The skill interacts with the user's home directory to manage configuration and store triage outputs (~/.claude/plugins/config/claude-for-legal/). These paths are specific to the tool's execution environment and represent standard application data persistence rather than an unauthorized access attempt.

fto-triage