The Agent Skills Directory

[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it aggregates untrusted user input (e.g., app descriptions, audience pain points, and competitive differentiators) and saves it into a shared file (app-marketing-context.md) intended for consumption by other skills.
Ingestion points: User-provided text for 'One-sentence description', 'User problems', and 'Key differentiators' in SKILL.md.
Boundary markers: The output format for app-marketing-context.md does not use delimiters or include instructions for downstream skills to ignore embedded commands within the data fields.
Capability inventory: While this skill primarily performs file writes, the resulting context file is explicitly designed to be read by tools like growth-analysis and metadata-optimization which may have expanded capabilities.
Sanitization: There is no evidence of input validation, escaping, or filtering for the data collected from the user or the get_app_detail tool.
[DATA_EXFILTRATION]: The skill actively requests and records sensitive business data, including estimated monthly revenue, download figures, and specific budget targets. Although the data is stored locally, the consolidation of sensitive performance metrics into a centralized, unencrypted plain-text document increases the risk of data exposure to any other skill or process operating within the agent's environment.

app-marketing-context