The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill implements dynamic loading of additional instructions using a computed path pattern: country-context/{country}/SKILL.md. The {country} variable is derived from user input. This pattern creates a risk where a malicious user could potentially influence the path to load unintended or unauthorized skill files if the platform does not strictly validate the input against an allowed list of countries or sanitise the path against traversal.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process free-form prose provided by a client (user input) through a 10-question intake process.
Ingestion points: The responses to Phase 1 questions defined in SKILL.md are directly incorporated into a 'Draft Project Brief'.
Boundary markers: The skill uses markdown headers and [TO CONFIRM] markers to delimit sections, but does not provide explicit technical delimiters to separate untrusted client prose from agent instructions.
Capability inventory: No evidence of shell execution, network exfiltration, or file-writing capabilities was found within this specific skill file.
Sanitization: There are no instructions provided to sanitize or escape special characters or instruction-like text within the user-provided responses before they are interpolated into the brief.

00-client-intake