The Agent Skills Directory

[DATA_EXFILTRATION]: The skill references the path ~/.config/growthbook/.env to manage the GB_API_KEY. This is a documented practice for GrowthBook integrations and is restricted to the vendor's configuration directory.
[COMMAND_EXECUTION]: The skill uses a bundled bash utility ${CLAUDE_PLUGIN_ROOT}/scripts/gb-call to perform specific GET requests against the GrowthBook API. The execution is limited to predefined workflow steps.
[EXTERNAL_DOWNLOADS]: The skill interacts with the GrowthBook API via official endpoints. As a well-known service provided by the skill's author (growthbook), these external interactions are considered safe and functional.
[PROMPT_INJECTION]: The skill ingests experiment data from the GrowthBook API to generate hypotheses. While this creates a surface for indirect prompt injection, it is mitigated by the fact that the data source is the user's own authenticated account, and the skill instructions mandate grounding all suggestions in specific historical evidence.
Ingestion points: Experiment metadata and results retrieved via gb-call in SKILL.md.
Boundary markers: None present.
Capability inventory: Limited to read-only API calls via gb-call; no write or administrative tool access.
Sanitization: Not applicable as the data is used for internal reasoning rather than command generation.

experiment-brainstorm