The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface because it processes content from external audit files to generate the final proposal.
Ingestion points: Audit data is loaded from ~/.geo-prospects/audits/<domain>*.md and prospect metadata from ~/.geo-prospects/prospects.json.
Boundary markers: No explicit delimiters or instructions to ignore embedded prompts are used when interpolating data into the proposal template.
Capability inventory: The skill uses Write and Bash tools to create proposal files and update local JSON records.
Sanitization: There is no evidence of input validation or content escaping for the audit findings before they are injected into the markdown template.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool for local filesystem operations, such as checking for the existence of audit files and managing directories within the ~/.geo-prospects/ path. This behavior is consistent with the skill's stated purpose of managing local business data.
[SAFE]: Data access is restricted to application-specific paths (~/.geo-prospects/) and does not attempt to access sensitive system files, credentials, or network resources for exfiltration.

geo-proposal