The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to scan for and analyze W-9 documentation. W-9 forms are highly sensitive as they contain full names, addresses, and Taxpayer Identification Numbers (Social Security Numbers or Employer Identification Numbers), posing a significant data exposure risk.
[DATA_EXFILTRATION]: The skill accesses the ~/.claude/projects/ directory to record telemetry. This directory is a sensitive area containing project context and history. Writing to this location outside the project root allows for persistent tracking and state retention across sessions.
[PROMPT_INJECTION]: The skill has a high attack surface for indirect prompt injection because it processes untrusted data from vendor databases, contract structures, and procurement workflows.
Ingestion points: Vendor databases, contract files, RFP/RFQ documents, and performance metrics.
Boundary markers: No delimiters or instructions to ignore embedded commands are present when processing these external data sources.
Capability inventory: The skill has file-write capabilities, allowing it to output analysis to docs/vendor-coordination-analysis.md and telemetry to ~/.claude/projects/.
Sanitization: There is no mention of escaping, validating, or sanitizing external content before it is processed or written to disk.

vendor-coordination