The Agent Skills Directory

[PROMPT_INJECTION]: The skill presents an attack surface for indirect prompt injection.
Ingestion points: The script scripts/explore.py reads interaction logs from ~/.opensquilla/logs/decisions-*.jsonl and code from tests/test_skills/router_fixtures/*.py.
Boundary markers: The skill does not implement boundary markers or explicit instructions to ignore commands within the logs.
Capability inventory: The skill can read local log files, aggregate data, and return results to the agent.
Sanitization: Log content is aggregated without specific sanitization for adversarial instructions.
[DATA_EXFILTRATION]: The skill accesses sensitive interaction history.
Evidence: The script reads from ~/.opensquilla/logs, which contains the agent's decision history.
Context: This is vendor-specific data access required for the skill's primary function; no external network transmission occurs.
[COMMAND_EXECUTION]: User-provided input is processed as a command-line argument.
Evidence: The query parameter is passed to the entrypoint script and handled by argparse in scripts/explore.py.
Context: Risk is mitigated by structured argument parsing and the platform's non-shell execution model.

history-explorer