The Agent Skills Directory

[COMMAND_EXECUTION]: The skill relies on the ae-cli tool to perform all data operations. It utilizes shell-native features like HEREDOCs (cat > /tmp/... <<'EOF') to generate temporary JSON configuration files and command substitution ($(cat ...)) to pass these definitions to the tool. This is a standard pattern for the agent's intended data analysis purpose.
[INDIRECT_PROMPT_INJECTION]: The skill contains an attack surface for indirect injection as it ingests external data (user IDs and report configurations) and interpolates them into command definitions.
Ingestion points: User-provided user_id (SKILL.md Step 2) and funnel report data retrieved via ae-cli analysis report get (SKILL.md Step 1).
Boundary markers: The skill uses JSON structures to wrap data but lacks explicit instructions for the agent to ignore potentially malicious control tokens within the ingested user IDs or report properties.
Capability inventory: The skill has extensive shell access to the ae-cli tool, including commands for adhoc queries, drilldown operations, and event detail exports.
Sanitization: There are no specific instructions to sanitize or validate the format of the user_id before it is placed into the JSON definition template.

funnel-misconceptions