The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides comprehensive documentation for executing mcpc CLI commands to interact with MCP servers. These instructions are intended for the user's manual testing and smoke-check workflows.
[CREDENTIALS_UNSAFE]: The skill documents commands that handle sensitive information, such as mcpc x402 import <private-key> and mcpc login with client secrets. It proactively includes security guardrails (Rule 13), warning that mcpc imports keys through positional arguments which may be exposed in process lists or shell history, and advises using throwaway wallets for testing.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect injection as it ingests data from external MCP servers via tool listings, resource reads, and prompt retrievals.
Ingestion points: External data enters through mcpc @session tools-list, resources-read, and prompts-get (referenced in references/guides/tool-resource-testing.md).
Boundary markers: The skill emphasizes the use of --json mode and jq for precise assertions, which acts as a boundary for machine-readable verification.
Capability inventory: The skill facilitates tools-call and resources-read operations.
Sanitization: The instructions recommend validating schemas using --schema and inspecting JSON payloads before relying on human-readable output.
[PROMPT_INJECTION]: A deterministic detector flagged concealment patterns due to descriptions of 'silent' tool behaviors (e.g., automatic profile injection or undocumented aliases). Analysis confirms these are benign descriptions of the underlying CLI tool's behavior and associated security flags (like --no-profile for anonymous testing) rather than instructions to deceive the user.

test-by-mcpc-cli