The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from the repository being integrated, such as README.md, CONTRIBUTING.md, and source code files, which could contain malicious instructions designed to influence the agent's behavior during the integration pipeline.
Ingestion points: The agent reads various repository files including README.md, CONTRIBUTING.md, entry points, and directory structures to comprehend the project's architecture as described in the pipeline.md file.
Boundary markers: The skill uses structured artifact files (goal.md, plan.md) to define the intent and mechanics of the integration, providing a level of separation between analysis and implementation.
Capability inventory: The skill can execute arbitrary shell commands for application testing (start, write_call, read_call defined in plan.md), modify source code files via a subagent implementation step, and perform network requests to fetch documentation from vendor-owned domains.
Sanitization: No explicit sanitization or filtering of the processed repository content is described before it is interpolated into agent prompts.
[COMMAND_EXECUTION]: The skill is designed to execute shell commands specified in an 'E2E recipe' within plan.md to verify the integration. These commands, which include launching the application and triggering specific read/write paths, are derived from the target repository's environment and configuration.
[EXTERNAL_DOWNLOADS]: The skill performs WebFetch operations to retrieve documentation, OpenAPI specifications, and additional skill definitions from vendor-controlled domains such as docs.mem0.ai, mcp.mem0.ai, and github.com/mem0ai. These resources are fetched as part of the integration process to ensure the agent uses up-to-date API information and official integration patterns.
[CREDENTIALS_UNSAFE]: The skill manages sensitive API keys (MEM0_API_KEY and OPENAI_API_KEY) required for the integration. It includes logic to retrieve these keys from environment variables or local vendor configuration files (e.g., ~/.mem0/config.json) and provides explicit instructions to avoid logging these secrets or committing them to version control, representing standard secret management practices for this vendor's tools.

mem0-integrate