The Agent Skills Directory

[SAFE]: The skill's logic is entirely dedicated to legal and HR expansion planning. It does not exhibit behaviors associated with credential theft, unauthorized network access, or malicious code execution.
[DATA_EXPOSURE_ANALYSIS]: The skill reads from and writes to specific files within a hidden configuration directory (~/.claude/plugins/config/claude-for-legal/).
Evidence: It loads CLAUDE.md and creates/updates expansion-[slug].yaml within that path.
Context: This activity is localized to the plugin's operational environment and is used for persistent state management rather than harvesting sensitive system or user data.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied data regarding staffing needs, which creates a potential surface for indirect prompt injection.
Ingestion points: Data enters the context via user responses in SKILL.md regarding job types, numbers, and timelines.
Boundary markers: No specific boundary markers are used to isolate untrusted input from the skill's analytical instructions.
Capability inventory: The skill is limited to reading and writing its own tracking files in SKILL.md; it has no network or shell execution capabilities.
Sanitization: No evidence of input sanitization or verification is present.
Assessment: Because the skill lacks dangerous capabilities like remote code execution or network exfiltration, the potential for harm from indirect injection is negligible.

expansion-kickoff