The Agent Skills Directory

[SAFE]: The skill follows security best practices for agent-cli interactions. It contains explicit instructions to never read, print, or persist sensitive access keys (AK/SK) or tokens in prompts or request artifacts (SKILL.md). It also leverages dry-run capabilities for contract validation before execution (best-practices.md).
[INDIRECT_PROMPT_INJECTION]: The skill handles log data which acts as a potential surface for indirect prompt injection, but the design and instructions mitigate this risk.
Ingestion points: Untrusted data enters the agent context through volclog tool exec log.search and volclog workflow exec log.ingest (SKILL.md, routing.md).
Boundary markers: The instructions mandate reading the full JSON envelope and using --jmes-filter for precise extraction, ensuring data is clearly separated from instructions (SKILL.md).
Capability inventory: All capabilities are limited to the volclog / ve CLI tool for log management and search; no dangerous tools like eval, exec, or arbitrary network operations are present (SKILL.md, sops.md).
Sanitization: The skill uses structured JSON via the --input flag, which enforces strict schema validation through the CLI tool's contract (SKILL.md, best-practices.md).

volcengine-tls