The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local shell commands using the Bash tool to perform diagnostics, including querying an internal SQLite database (sessions.db) and running project-local JavaScript files (validate-registry.mjs, fired-census.mjs). These actions are scoped to the project's observability framework and are performed in a read-only manner.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from local telemetry logs and state files to report on system health. While these logs may contain information derived from external tool outputs, the skill's restricted read operations (limited to file metadata and the first 10 lines) and diagnostic context minimize the potential for prompt injection attacks. No unsafe use of this data was identified.
[SAFE]: The skill references a coordination layer and telemetry sink owned by the vendor (yonatangross/OrchestKit). These are documented infrastructure components, and the skill performs no unauthorized network operations or external data transfers.

telemetry-inspect