The Agent Skills Directory

[PERSISTENCE]: The references/onboarding.md file provides explicit instructions to modify shell configuration files (~/.bashrc, ~/.zshrc) and use the setx command on Windows to persist the LINKFOX_AGENT_API_KEY environment variable.
[CREDENTIALS_UNSAFE]: The scripts/onboarding.py script prints the retrieved api_key to standard output in a JSON format upon successful login. This allows the AI agent to access and potentially leak the sensitive credential. Additionally, the script contains a hardcoded fallback identifier string in the _LOGIN_FIXED_UID variable.
[COMMAND_EXECUTION]: The skill relies on the execution of local Python scripts (scripts/etsy_store_query.py and scripts/onboarding.py) through shell commands, using JSON strings as command-line arguments.
[EXTERNAL_DOWNLOADS]: The scripts/onboarding.py script prompts the user to download and install external Python packages (requests, qrcode, pillow) via pip to enable specific functionalities.
[INDIRECT_PROMPT_INJECTION]: The skill is vulnerable to indirect prompt injection because it ingests untrusted user data—such as search keywords, phone numbers, and authentication codes—which are then passed to internal scripts and external API endpoints.
Ingestion points: User-provided query parameters in references/api.md and authentication inputs in scripts/onboarding.py.
Boundary markers: None identified.
Capability inventory: Execution of Python scripts, file system writes to the session directory, and network access to Linkfox vendor domains.
Sanitization: The scripts use json.loads and argparse for basic input parsing, but do not implement rigorous sanitization to prevent adversarial instructions embedded in data from influencing agent behavior.

linkfox-etsy-store-query