The Agent Skills Directory

[DATA_EXFILTRATION]: The skill instructs the agent to read sensitive local configuration files, specifically the Contentstack MCP OAuth session data at ~/Library/Application Support/ContentstackMCP/oauth-config.json (and its platform equivalents). These files store OAuth access and refresh tokens which are used for authenticating management API calls.
[DYNAMIC_EXECUTION]: The skill frequently uses python3 -c and multi-line Python scripts passed via stdin (e.g., in author-composition-encoding.md and match-existing-pattern.md) to perform data processing tasks such as JSON manipulation, zlib compression, and tree analysis. These scripts are constructed at runtime and executed in the local environment.
[INDIRECT_PROMPT_INJECTION]: The skill has a large attack surface as it regularly scans untrusted project files like package.json, lockfiles, and source code files (.tsx, .js, etc.) to classify project architecture and components. This data is used to drive subsequent logic and command generation.
Ingestion points: Project configuration and source files are read in analyze-project-fit.md, discover-sections.md, and decompose-jsx-to-atomics.md.
Boundary markers: The skill uses markdown headers and instructional markers but does not explicitly document the sanitization of data retrieved from these project files.
Capability inventory: The agent possesses capabilities for shell command execution (npm, ls, grep, cat), local file access, and network operations via curl.
Sanitization: No specific methods for escaping or validating project-derived strings before interpolation into shell commands are detailed.
[DYNAMIC_CONTEXT_INJECTION]: The skill makes use of the !command syntax within its documentation (e.g., in SKILL.md) to provide the agent with interactive commands intended to be given to the user for execution.
[COMMAND_EXECUTION]: The skill performs extensive inspection of the local environment using commands like npm ls, ls, grep, and cat. It also provides instructions for the user to install packages and perform OAuth authentication flows.
[EXTERNAL_DOWNLOADS]: The skill facilitates the installation of various Contentstack SDKs and tools from the public npm registry. All detected network operations and downloads target Contentstack's official domains or package names, which are associated with the skill's primary vendor.

studio