The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to access sensitive authentication information stored in specific local files, notably ~/.config/environment.d/secrets.conf for the ZAI_AUTH_TOKEN, and references the use of environment variables like MINIMAX_CODING_KEY for server authentication.
[REMOTE_CODE_EXECUTION]: The skill recommends executing code from remote sources at runtime using package runners. Evidence includes instructions to use uvx minimax-coding-plan-mcp and npx @z_ai/mcp-server to initialize MCP server sessions.
[COMMAND_EXECUTION]: The skill is centered on instructing the agent to generate and execute shell commands using the mcpc CLI tool for interacting with MCP servers, which includes managing sessions and calling arbitrary tools.
[DATA_EXFILTRATION]: The skill demonstrates patterns for passing sensitive bearer tokens to external URLs via command-line arguments, such as mcpc connect mcp.example.com @api --header "Authorization: Bearer $TOKEN".
[INDIRECT_PROMPT_INJECTION]: The skill establishes a workflow for processing untrusted data from external sources, which constitutes an injection attack surface.
Ingestion points: Tools like web_search_prime, webReader, and searxng_web_search ingest external internet content into the agent's context.
Boundary markers: There are no instructions for using delimiters or warnings to ignore instructions embedded in the external content fetched by tools.
Capability inventory: The agent is granted shell-level access via the mcpc CLI, allowing for network communication and access to local configuration files.
Sanitization: The skill lacks instructions for sanitizing or validating tool outputs before the agent processes them.

mcpc