The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill builds SQL queries by performing manual string substitution on the contact-insert.sql template using data extracted from business cards and web searches. While it instructs the agent to escape single quotes, this pattern is inherently risky and could lead to SQL injection if the escaping is bypassed or incomplete.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from business card images via OCR and from external web search results. This data is then passed to the sales-follow-up-mail sub-agent to generate email drafts. A malicious business card or search result could contain instructions designed to manipulate the agent's output or behavior.
Ingestion points: OCR extraction from card images in Step 2 and WebSearch results in Step 5 of SKILL.md.
Boundary markers: Absent. The instructions do not provide delimiters or warnings to the sub-agent to ignore instructions contained within the external data.
Capability inventory: Database execution via mcp__supabase__execute_sql and email operations via mcp__gmail__draft_email and mcp__gmail__send_email as defined in SKILL.md.
Sanitization: Limited to single-quote doubling for SQL inputs; no sanitization is performed on data passed to the email drafting sub-agent.
[EXTERNAL_DOWNLOADS]: The skill is installed from a remote GitHub repository (github.com/peter-tu-zynkr/zynkr-skill-builder). This source belongs to the skill's author.

sales-specialist