The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The search functionality in scripts/search.sh fetches charity metadata (names, descriptions) from an external API (api.endaoment.org) and displays it to the agent. This untrusted data could potentially contain malicious instructions.
Ingestion points: scripts/search.sh API responses.
Boundary markers: None identified.
Capability inventory: The skill can execute blockchain transactions via bankr agent in scripts/donate.sh.
Sanitization: The description field is truncated to 200 characters, reducing the available surface for long-form injection attacks.
[COMMAND_EXECUTION]: The scripts/donate.sh script dynamically generates transaction data and instructs the agent to execute it using the bankr agent command. While this is the primary intended function of the skill, it represents a high-privilege interaction where the script controls the agent's blockchain activity.

endaoment