The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes a local Python script (scripts/check_freshness.py) for verifying package versions and documentation links. The agent is instructed to execute this script to ensure that generated integration plans are based on current dependency data.
[EXTERNAL_DOWNLOADS]: During the freshness verification phase, the skill performs network requests to the NPM registry and GitHub API to retrieve version metadata. It also probes documentation URLs found within its own files to verify link liveness. All network activity targets well-known services or official documentation repositories associated with the Starknet ecosystem.
[PROMPT_INJECTION]: The skill scans local repository files to detect the tech stack, creating a surface for indirect prompt injection from repository content.
Ingestion points: Repository configuration files (e.g., package.json, Scarb.toml) and Cairo source code are read to identify the project stack.
Boundary markers: The skill requires explicit human review and approval of the generated integration plan (STRK20_INTEGRATION_PLAN.md) before any code execution or modification occurs.
Capability inventory: The skill can install NPM packages, modify application source code, and configure transaction layers.
Sanitization: The skill utilizes templated plan generation and enforces specific version pins from its own verified reference files.
[SAFE]: The skill exclusively utilizes resources and dependencies from trusted organizations and well-known services. It incorporates safety guardrails, including human-in-the-loop checkpoints for all execution phases and explicit warnings against handling sensitive cryptographic material like viewing keys.

strk20-privacy-integration