The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the Barker API, creating a potential surface for indirect prompt injection if the external service provides malicious instructions in its response fields.
Ingestion points: Data is fetched from api.barker.money endpoints for market overview and DeFi vaults as described in SKILL.md and SUMMARY.md.
Boundary markers: The skill includes a dedicated "Security: External Data Boundary" section in SKILL.md and SUMMARY.md that explicitly instructs the agent to separate data from instructions.
Capability inventory: The agent has capabilities to execute shell commands (for version checks) and perform network requests.
Sanitization: The instructions explicitly direct the assistant to render values as data and avoid executing any imperative text found in API responses.
[EXTERNAL_DOWNLOADS]: The skill includes a pre-flight version check that fetches metadata from a remote repository.
Evidence: The bash script in SKILL.md uses curl to fetch plugin.yaml from the official GitHub repository to verify the local version against the latest available update.
[COMMAND_EXECUTION]: The skill provides a bash script for the agent to run, which manages a local version cache.
Evidence: SKILL.md contains a script that uses standard shell utilities to manage a cache file located within the user's home directory.

stablecoin-chain-explorer