The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The script scripts/ensure-daily-skill-update.mjs is designed to run automatically and executes npx -y skills@latest update, which downloads and runs the latest version of an unpinned package from the public npm registry.
[PRIVILEGE_ESCALATION]: The update mechanism in scripts/ensure-daily-skill-update.mjs uses the -g flag for global installation, which attempts to modify system-level directories and typically requires elevated permissions.
[EXTERNAL_DOWNLOADS]: The skill initiates network requests to download remote packages via npx and fetches document content from external URLs provided in work item metadata using scripts/handoff_gate.py.
[COMMAND_EXECUTION]: Extensive use of subprocess.run is found in scripts/validate_release_change_coverage.py and scripts/yunxiao_cli_runtime.py to execute git and the aliyun CLI for cloud infrastructure and source code management.
[DYNAMIC_EXECUTION]: The skill contains logic in scripts/yunxiao_cli_gateway.py to dynamically generate and create new cloud pipeline definitions (flow-create-pipeline) by parsing and modifying existing YAML configurations at runtime.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data which creates an attack surface for indirect prompt injection.
Ingestion points: Commit histories, work item descriptions, and merge request metadata are ingested via the aliyun CLI in scripts/validate_release_change_coverage.py and scripts/release_description.md.
Boundary markers: The skill uses HTML comment delimiters such as  to separate managed data within work item descriptions.
Capability inventory: The agent can merge source code, trigger production pipelines, and update work item statuses through the yunxiao_cli_gateway.py script.
Sanitization: The skill implements a scrub function in scripts/yunxiao_cli_runtime.py to redact tokens and uses strict JSON schema validation for release evidence.

yunxiao-release-operations