The Agent Skills Directory

[COMMAND_EXECUTION]: The diagnostic script doctor.sh and associated playbooks execute various system commands including systemctl for service monitoring, sqlite3 for ledger inspection, gsettings for desktop environment checks, and mc for MinIO storage administration.
[DYNAMIC_EXECUTION]: The skill uses python3 heredocs and inline scripts to dynamically process JSON state data, verify library imports (such as torch and nemo), and execute CUDA smoke tests to ensure diarization functionality.
[DATA_EXFILTRATION]: The skill resolves live configuration by reading from the /proc filesystem (e.g., /proc/pid/environ and /proc/pid/status) to identify active roots and ledger paths, and uses the op CLI to verify that required secrets are resolvable from 1Password without logging their values.
[EXTERNAL_DOWNLOADS]: The skill performs connectivity checks via curl to external LLM provider endpoints (OpenRouter, Pokee) and MinIO storage services to verify availability and model presence.
[INDIRECT_PROMPT_INJECTION]: The skill processes data from the local pipeline ledger and transcript files during its diagnostic routines.
Ingestion points: Data is pulled from the wax.db SQLite database and state.json mirror.
Boundary markers: No specific delimiters or safety instructions are used when reading internal state.
Capability inventory: The agent can execute system commands, restart services, and perform network requests.
Sanitization: Diagnostic checks assume the integrity of the local ledger data without secondary sanitization.

heyma-pipeline-doctor