The Agent Skills Directory

[DATA_EXFILTRATION]: The skill workflow includes instructions to explicitly unset user-configured privacy flags. It uses env -u DISABLE_TELEMETRY -u DO_NOT_TRACK when running verification commands to force telemetry data to be sent to the ecosystem backend, bypassing user-defined tracking opt-outs.\n- [EXTERNAL_DOWNLOADS]: The skill fetches and executes the latest version of the skills utility from the NPM registry via npx skills@latest. It also relies on the gh skill extension for the GitHub CLI.\n- [COMMAND_EXECUTION]: The skill executes a local Python script (scripts/validate_skills.py) to validate repository content during the pre-flight phase.\n- [PRIVILEGE_ESCALATION]: The skill performs high-impact administrative actions on the GitHub repository, such as flipping repository visibility from private to public and configuring security rulesets and private vulnerability reporting (PVR).\n- [INDIRECT_PROMPT_INJECTION]: The skill processes repository contents and configuration files which may contain untrusted data, while possessing broad administrative and execution capabilities.\n
Ingestion points: Reads files within the skills/ directory and the skills.sh.json configuration file.\n
Boundary markers: No delimiters or explicit instructions are provided to the agent to ignore potentially malicious content within these files.\n
Capability inventory: Administrative GitHub repository access, package execution, and local script execution.\n
Sanitization: No sanitization of the content extracted from skill files is performed before processing.

publish-repo