The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructs the agent to run local Python scripts (e.g., scout_competitors.py) to manage the discovery and ranking process. These scripts interact with the local file system to store and organize results.
[EXTERNAL_DOWNLOADS]: The skill performs automated HTTP requests to itunes.apple.com, apps.apple.com, and play.google.com to fetch app metadata and rankings.
[PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection as it ingests untrusted text (app names, developer names, genres) from public app stores.
Ingestion points: Application metadata fetched via AppleClient and GooglePlayClient in discovery.py.
Boundary markers: None. The skill renders untrusted data directly into Markdown tables in shortlist.md without delimiters or instructions to the agent to ignore embedded commands.
Capability inventory: The skill generates lists of URLs (review-urls.txt) intended for subsequent processing by other tools and writes JSON/Markdown reports to the workspace.
Sanitization: Content is only sanitized for Markdown table syntax (escaping pipes); there is no filtering for natural language instructions.
[CREDENTIALS_UNSAFE]: In apple_store.py, the AppleClient class implements an --insecure flag that invokes ssl._create_unverified_context(). This disables TLS certificate validation, exposing the agent's network traffic to potential man-in-the-middle attacks.

copycat-scout