The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell scripts (resolve-canister-id.sh and fetch-candid.sh) which invoke system utilities including curl, python3, and the icp command-line interface to perform its primary functions.
[EXTERNAL_DOWNLOADS]: Connects to the Internet Computer Dashboard API (ic-api.internetcomputer.org) to resolve canister names and retrieves canister metadata from the Internet Computer network.
[PROMPT_INJECTION]: The skill processes Candid interface definitions fetched from external canisters. This creates an indirect prompt injection surface where a canister owner could include malicious instructions within method names or documentation to influence the agent's summary or subsequent actions.
Ingestion points: API responses in resolve-canister-id.sh and Candid files downloaded in fetch-candid.sh.
Boundary markers: Absent; the agent reads the retrieved Candid files and generates a summary based on the raw content.
Capability inventory: Shell script execution via Bash and file system access via Read.
Sanitization: No sanitization or escaping is applied to the Candid interface content before it is presented to the agent for analysis.

canhelp