The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions in SKILL.md direct the agent to execute a bundled Python utility scripts/check_freshness.py to ensure that version numbers and documentation links remain accurate relative to the live protocol status.
[EXTERNAL_DOWNLOADS]: The bundled freshness script performs network requests to several external domains to validate protocol metadata and documentation availability. Targeted domains include registry.npmjs.org, raw.githubusercontent.com (for starkware-libs and starknet-specs), strk20-by-example.org, and strk20.starknet.io. These are read-only status checks against well-known ecosystem resources.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of external data from protocol websites to update the agent's internal state.
Ingestion points: Output of the scripts/check_freshness.py script, which fetches versioning and liveness data from external URLs.
Boundary markers: Absent from the script output processing instructions.
Capability inventory: The skill allows for documentation retrieval and routing advice; the utility script includes network operations (urllib.request) but no file-write or code-execution capabilities.
Sanitization: The script uses standard JSON parsing and regex for fact extraction, providing a low-risk interface for factual updates.

strk20-privacy