The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from an external JSON file located at ~/.claude-marketing/brands/{slug}/campaigns/{campaign_id}/plan.json.
Ingestion points: Campaign plan JSON file (plan.json).
Boundary markers: The skill implements a mandatory 'dry-run' preview and requires the user to explicitly type 'yes' before any actions are executed, which serves as a significant security gate.
Capability inventory: The skill uses the Bash tool to execute Python scripts and curl commands, and has file read/write permissions.
Sanitization: The instructions do not specify any explicit sanitization or escaping of the ingested JSON fields (e.g., campaign names, IDs, or URLs) before they are interpolated into shell command strings.
[COMMAND_EXECUTION]: The skill constructs and executes multiple shell commands (e.g., curl and various python scripts) using variables derived from the ingested campaign plan. The interpolation of this external data directly into shell strings (e.g., curl -sS -o /dev/null -w "%{http_code}" "{landing_url}") creates a potential command injection surface if the source file contains shell metacharacters, though the risk is mitigated by the manual user confirmation step.

launch-campaign